3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1999, 2000,
4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2012 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
11 /* IMPORTANT NOTE: Everything whose name begins with an underscore is for
12 * internal core Perl use only. */
14 #ifndef PERL_HANDY_H_ /* Guard against nested #inclusion */
18 # define Null(type) ((type)NULL)
21 =for apidoc_section $string
22 =for apidoc AmnU||Nullch
23 Null character pointer. (No longer available when C<PERL_CORE> is
26 =for apidoc_section $SV
27 =for apidoc AmnU||Nullsv
28 Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
32 Below are signatures of functions from config.h which can't easily be gleaned
33 from it, and are very unlikely to change
35 =for apidoc_section $signals
36 =for apidoc Am|int|Sigsetjmp|jmp_buf env|int savesigs
37 =for apidoc Am|void|Siglongjmp|jmp_buf env|int val
39 =for apidoc_section $filesystem
40 =for apidoc Am|void *|FILE_ptr|FILE * f
41 =for apidoc Am|Size_t|FILE_cnt|FILE * f
42 =for apidoc Am|void *|FILE_base|FILE * f
43 =for apidoc Am|Size_t|FILE_bufsiz|FILE *f
45 =for apidoc_section $string
46 =for apidoc Amu|token|CAT2|token x|token y
47 =for apidoc Amu|string|STRINGIFY|token x
49 =for apidoc_section $numeric
50 =for apidoc Am|double|Drand01
51 =for apidoc Am|void|seedDrand01|Rand_seed_t x
52 =for apidoc Am|char *|Gconvert|double x|Size_t n|bool t|char * b
57 # define Nullch Null(char*)
58 # define Nullfp Null(PerlIO*)
59 # define Nullsv Null(SV*)
72 =for apidoc_section $SV
73 =for apidoc Am |AV * |MUTABLE_AV |AV * p
74 =for apidoc_item |CV * |MUTABLE_CV |CV * p
75 =for apidoc_item |GV * |MUTABLE_GV |GV * p
76 =for apidoc_item |HV * |MUTABLE_HV |HV * p
77 =for apidoc_item |IO * |MUTABLE_IO |IO * p
78 =for apidoc_item |void *|MUTABLE_PTR|void * p
79 =for apidoc_item |SV * |MUTABLE_SV |SV * p
81 The C<MUTABLE_I<*>>() macros cast pointers to the types shown, in such a way
82 (compiler permitting) that casting away const-ness will give a warning;
86 AV *av1 = (AV*)sv; <== BAD: the const has been silently
88 AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
90 C<MUTABLE_PTR> is the base macro used to derive new casts. The other
91 already-built-in ones return pointers to what their names indicate.
95 The brace group version will raise a diagnostic if 'p' is const; the other
96 blindly casts away const.
98 #if defined(PERL_USE_GCC_BRACE_GROUPS)
99 # define MUTABLE_PTR(p) ({ void *p_ = (p); p_; })
101 # define MUTABLE_PTR(p) ((void *) (p))
104 #define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
105 #define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
106 #define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
107 #define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
108 #define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
109 #define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
112 # include <stdbool.h>
116 =for apidoc_section $casting
117 =for apidoc Am|bool|cBOOL|bool expr
119 Cast-to-bool. When Perl was able to be compiled on pre-C99 compilers, a
120 C<(bool)> cast didn't necessarily do the right thing, so this macro was
121 created (and made somewhat complicated to work around bugs in old
122 compilers). Now, many years later, and C99 is used, this is no longer
123 required, but is kept for backwards compatibility.
127 #define cBOOL(cbool) ((bool) (cbool))
129 /* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
130 * XXX Should really be a Configure probe, with HAS__FUNCTION__
131 * and FUNCTION__ as results.
132 * XXX Similarly, a Configure probe for __FILE__ and __LINE__ is needed. */
133 #if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
134 # define FUNCTION__ __func__
135 #elif (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tur64, -c99 not known, only -std1). */
136 # define FUNCTION__ ""
138 # define FUNCTION__ __FUNCTION__ /* Common extension. */
141 /* XXX A note on the perl source internal type system. The
142 original intent was that I32 be *exactly* 32 bits.
144 Currently, we only guarantee that I32 is *at least* 32 bits.
145 Specifically, if int is 64 bits, then so is I32. (This is the case
146 for the Cray.) This has the advantage of meshing nicely with
147 standard library calls (where we pass an I32 and the library is
148 expecting an int), but the disadvantage that an I32 is not 32 bits.
149 Andy Dougherty August 1996
151 There is no guarantee that there is *any* integral type with
152 exactly 32 bits. It is perfectly legal for a system to have
153 sizeof(short) == sizeof(int) == sizeof(long) == 8.
155 Similarly, there is no guarantee that I16 and U16 have exactly 16
158 For dealing with issues that may arise from various 32/64-bit
159 systems, we will ask Configure to check out
161 SHORTSIZE == sizeof(short)
162 INTSIZE == sizeof(int)
163 LONGSIZE == sizeof(long)
164 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
165 PTRSIZE == sizeof(void *)
166 DOUBLESIZE == sizeof(double)
167 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
171 #ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
172 # include <inttypes.h>
173 # ifdef INT32_MIN_BROKEN
175 # define INT32_MIN (-2147483647-1)
177 # ifdef INT64_MIN_BROKEN
179 # define INT64_MIN (-9223372036854775807LL-1)
195 /* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
196 Please search CHAR_MAX in perl.h for further details. */
198 # define U8_MAX UINT8_MAX
200 # define U8_MAX PERL_UCHAR_MAX
203 # define U8_MIN UINT8_MIN
205 # define U8_MIN PERL_UCHAR_MIN
209 # define I16_MAX INT16_MAX
211 # define I16_MAX PERL_SHORT_MAX
214 # define I16_MIN INT16_MIN
216 # define I16_MIN PERL_SHORT_MIN
219 # define U16_MAX UINT16_MAX
221 # define U16_MAX PERL_USHORT_MAX
224 # define U16_MIN UINT16_MIN
226 # define U16_MIN PERL_USHORT_MIN
230 # define I32_MAX INT32_MAX
232 # define I32_MAX PERL_INT_MAX
234 # define I32_MAX PERL_LONG_MAX
237 # define I32_MIN INT32_MIN
239 # define I32_MIN PERL_INT_MIN
241 # define I32_MIN PERL_LONG_MIN
244 # ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
245 # define U32_MAX UINT_MAX
247 # define U32_MAX 4294967295U
250 # define U32_MAX PERL_UINT_MAX
252 # define U32_MAX PERL_ULONG_MAX
255 # define U32_MIN UINT32_MIN
257 # define U32_MIN PERL_UINT_MIN
259 # define U32_MIN PERL_ULONG_MIN
263 =for apidoc_section $integer
264 =for apidoc Ay|| PERL_INT_FAST8_T
265 =for apidoc_item PERL_INT_FAST16_T
266 =for apidoc_item PERL_UINT_FAST8_T
267 =for apidoc_item PERL_UINT_FAST16_T
269 These are equivalent to the correspondingly-named C99 typedefs on platforms
270 that have those; they evaluate to C<int> and C<unsigned int> on platforms that
271 don't, so that you can portably take advantage of this C99 feature.
276 typedef int_fast8_t PERL_INT_FAST8_T;
277 typedef uint_fast8_t PERL_UINT_FAST8_T;
278 typedef int_fast16_t PERL_INT_FAST16_T;
279 typedef uint_fast16_t PERL_UINT_FAST16_T;
281 typedef int PERL_INT_FAST8_T;
282 typedef unsigned int PERL_UINT_FAST8_T;
283 typedef int PERL_INT_FAST16_T;
284 typedef unsigned int PERL_UINT_FAST16_T;
287 /* log(2) (i.e., log base 10 of 2) is pretty close to 0.30103, just in case
288 * anyone is grepping for it. So BIT_DIGITS gives the number of decimal digits
289 * required to represent any possible unsigned number containing N bits.
290 * TYPE_DIGITS gives the number of decimal digits required to represent any
291 * possible unsigned number of type T. */
292 #define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log10(2) =~ 146/485 */
293 #define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
294 #define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
296 /* Unused by core; should be deprecated */
297 #define Ctl(ch) ((ch) & 037)
299 #if defined(PERL_CORE) || defined(PERL_EXT)
301 # define MIN(a,b) ((a) < (b) ? (a) : (b))
304 # define MAX(a,b) ((a) > (b) ? (a) : (b))
308 /* Returns a boolean as to whether the input unsigned number is a power of 2
309 * (2**0, 2**1, etc). In other words if it has just a single bit set.
310 * If not, subtracting 1 would leave the uppermost bit set, so the & would
312 #if defined(PERL_CORE) || defined(PERL_EXT)
313 # define isPOWER_OF_2(n) ((n) && ((n) & ((n)-1)) == 0)
316 /* Returns a mask with the lowest n bits set */
317 #define nBIT_MASK(n) ((UINTMAX_C(1) << (n)) - 1)
319 /* The largest unsigned number that will fit into n bits */
320 #define nBIT_UMAX(n) nBIT_MASK(n)
323 =for apidoc_section $directives
324 =for apidoc Am||__ASSERT_|bool expr
326 This is a helper macro to avoid preprocessor issues, replaced by nothing
327 unless under DEBUGGING, where it expands to an assert of its argument,
328 followed by a comma (hence the comma operator). If we just used a straight
329 assert(), we would get a comma with nothing before it when not DEBUGGING.
333 We also use empty definition under Coverity since the __ASSERT_
334 checks often check for things that Really Cannot Happen, and Coverity
335 detects that and gets all excited. */
337 #if defined(DEBUGGING) && !defined(__COVERITY__) \
338 && ! defined(PERL_SMALL_MACRO_BUFFER)
339 # define __ASSERT_(statement) assert(statement),
341 # define __ASSERT_(statement)
345 =for apidoc_section $SV
347 =for apidoc Ama|SV*|newSVpvs|"literal string"
348 Like C<newSVpvn>, but takes a literal string instead of a
351 =for apidoc Ama|SV*|newSVpvs_flags|"literal string"|U32 flags
352 Like C<newSVpvn_flags>, but takes a literal string instead of
353 a string/length pair.
355 =for apidoc Ama|SV*|newSVpvs_share|"literal string"
356 Like C<newSVpvn_share>, but takes a literal string instead of
357 a string/length pair and omits the hash parameter.
359 =for apidoc Am|void|sv_catpvs_flags|SV* sv|"literal string"|I32 flags
360 Like C<sv_catpvn_flags>, but takes a literal string instead
361 of a string/length pair.
363 =for apidoc Am|void|sv_catpvs_nomg|SV* sv|"literal string"
364 Like C<sv_catpvn_nomg>, but takes a literal string instead of
365 a string/length pair.
367 =for apidoc Am|void|sv_catpvs|SV* sv|"literal string"
368 Like C<sv_catpvn>, but takes a literal string instead of a
371 =for apidoc Am|void|sv_catpvs_mg|SV* sv|"literal string"
372 Like C<sv_catpvn_mg>, but takes a literal string instead of a
375 =for apidoc Am|SV *|sv_setref_pvs|SV *const rv|const char *const classname|"literal string"
376 Like C<sv_setref_pvn>, but takes a literal string instead of
377 a string/length pair.
379 =for apidoc_section $string
381 =for apidoc Ama|char*|savepvs|"literal string"
382 Like C<savepvn>, but takes a literal string instead of a
385 =for apidoc Ama|char*|savesharedpvs|"literal string"
386 A version of C<savepvs()> which allocates the duplicate string in memory
387 which is shared between threads.
389 =for apidoc_section $GV
391 =for apidoc Am|HV*|gv_stashpvs|"name"|I32 create
392 Like C<gv_stashpvn>, but takes a literal string instead of a
395 =for apidoc_section $HV
397 =for apidoc Am|SV**|hv_fetchs|HV* tb|"key"|I32 lval
398 Like C<hv_fetch>, but takes a literal string instead of a
400 =for apidoc_section $lexer
402 =for apidoc Amx|void|lex_stuff_pvs|"pv"|U32 flags
404 Like L</lex_stuff_pvn>, but takes a literal string instead of
405 a string/length pair.
410 #define ASSERT_IS_LITERAL(s) ("" s "")
413 =for apidoc_section $string
415 =for apidoc Amu|pair|STR_WITH_LEN|"literal string"
417 Returns two comma separated tokens of the input literal string, and its length.
418 This is convenience macro which helps out in some API calls.
419 Note that it can't be used as an argument to macros or functions that under
420 some configurations might be macros, which means that it requires the full
421 Perl_xxx(aTHX_ ...) form for any API calls where it's used.
426 #define STR_WITH_LEN(s) ASSERT_IS_LITERAL(s), (sizeof(s)-1)
428 /* STR_WITH_LEN() shortcuts */
429 #define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
430 #define newSVpvs_flags(str,flags) \
431 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
432 #define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
433 #define sv_catpvs_flags(sv, str, flags) \
434 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
435 #define sv_catpvs_nomg(sv, str) \
436 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
437 #define sv_catpvs(sv, str) \
438 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
439 #define sv_catpvs_mg(sv, str) \
440 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
441 #define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
442 #define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
443 #define sv_setref_pvs(rv, classname, str) \
444 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
445 #define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
446 #define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
447 #define gv_stashpvs(str, create) \
448 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
450 #define gv_fetchpvs(namebeg, flags, sv_type) \
451 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), flags, sv_type)
452 #define gv_fetchpvn gv_fetchpvn_flags
453 #define sv_catxmlpvs(dsv, str, utf8) \
454 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
457 #define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
459 #define get_cvs(str, flags) \
460 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
462 /* internal helpers */
464 #ifndef PERL_VERSION_MAJOR
465 # define PERL_VERSION_MAJOR PERL_REVISION
467 # undef PERL_REVISION /* We don't want code to be using these */
469 #ifndef PERL_VERSION_MINOR
470 # define PERL_VERSION_MINOR PERL_VERSION
474 #ifndef PERL_VERSION_PATCH
475 # define PERL_VERSION_PATCH PERL_SUBVERSION
477 # undef PERL_SUBVERSION
480 #define PERL_JNP_TO_DECIMAL_(maJor,miNor,Patch) \
481 /* '10*' leaves room for things like alpha, beta, releases */ \
482 (10 * ((maJor) * 1000000) + ((miNor) * 1000) + (Patch))
483 #define PERL_DECIMAL_VERSION_ \
484 PERL_JNP_TO_DECIMAL_(PERL_VERSION_MAJOR, PERL_VERSION_MINOR, \
488 =for apidoc_section $versioning
489 =for apidoc AmR|bool|PERL_VERSION_EQ|const U8 major|const U8 minor|const U8 patch
490 =for apidoc_item PERL_VERSION_GE
491 =for apidoc_item PERL_VERSION_GT
492 =for apidoc_item PERL_VERSION_LE
493 =for apidoc_item PERL_VERSION_LT
494 =for apidoc_item PERL_VERSION_NE
496 Returns whether or not the perl currently being compiled has the specified
497 relationship to the perl given by the parameters. For example,
499 #if PERL_VERSION_GT(5,24,2)
500 code that will only be compiled on perls after v5.24.2
505 Note that this is usable in making compile-time decisions
507 You may use the special value '*' for the final number to mean ALL possible
510 #if PERL_VERSION_EQ(5,31,'*')
512 means all perls in the 5.31 series. And
514 #if PERL_VERSION_NE(5,24,'*')
516 means all perls EXCEPT 5.24 ones. And
518 #if PERL_VERSION_LE(5,9,'*')
522 #if PERL_VERSION_LT(5,10,0)
524 This means you don't have to think so much when converting from the existing
525 deprecated C<PERL_VERSION> to using this macro:
527 #if PERL_VERSION <= 9
531 #if PERL_VERSION_LE(5,9,'*')
536 /* N.B. These don't work if the patch version is 42 or 92, as those are what
537 * '*' is in ASCII and EBCDIC respectively */
538 # define PERL_VERSION_EQ(j,n,p) \
540 ? ( (j) == PERL_VERSION_MAJOR \
541 && (n) == PERL_VERSION_MINOR) \
542 : (PERL_DECIMAL_VERSION_ == PERL_JNP_TO_DECIMAL_(j,n,p)))
543 # define PERL_VERSION_NE(j,n,p) (! PERL_VERSION_EQ(j,n,p))
545 # define PERL_VERSION_LT(j,n,p) /* < '*' effectively means < 0 */ \
546 (PERL_DECIMAL_VERSION_ < PERL_JNP_TO_DECIMAL_( (j), \
548 (((p) == '*') ? 0 : p)))
549 # define PERL_VERSION_GE(j,n,p) (! PERL_VERSION_LT(j,n,p))
551 # define PERL_VERSION_LE(j,n,p) /* <= '*' effectively means < n+1 */ \
552 (PERL_DECIMAL_VERSION_ < PERL_JNP_TO_DECIMAL_( (j), \
553 (((p) == '*') ? ((n)+1) : (n)), \
554 (((p) == '*') ? 0 : p)))
555 # define PERL_VERSION_GT(j,n,p) (! PERL_VERSION_LE(j,n,p))
558 =for apidoc_section $string
560 =for apidoc Am|bool|strNE|char* s1|char* s2
561 Test two C<NUL>-terminated strings to see if they are different. Returns true
564 =for apidoc Am|bool|strEQ|char* s1|char* s2
565 Test two C<NUL>-terminated strings to see if they are equal. Returns true or
568 =for apidoc Am|bool|strLT|char* s1|char* s2
569 Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
570 second, C<s2>. Returns true or false.
572 =for apidoc Am|bool|strLE|char* s1|char* s2
573 Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
574 equal to the second, C<s2>. Returns true or false.
576 =for apidoc Am|bool|strGT|char* s1|char* s2
577 Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
578 the second, C<s2>. Returns true or false.
580 =for apidoc Am|bool|strGE|char* s1|char* s2
581 Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
582 or equal to the second, C<s2>. Returns true or false.
584 =for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
585 Test two C<NUL>-terminated strings to see if they are different. The C<len>
586 parameter indicates the number of bytes to compare. Returns true or false. (A
587 wrapper for C<strncmp>).
589 =for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
590 Test two C<NUL>-terminated strings to see if they are equal. The C<len>
591 parameter indicates the number of bytes to compare. Returns true or false. (A
592 wrapper for C<strncmp>).
594 =for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
595 Test two buffers (which may contain embedded C<NUL> characters, to see if they
596 are equal. The C<len> parameter indicates the number of bytes to compare.
597 Returns true or false. It is undefined behavior if either of the buffers
598 doesn't contain at least C<len> bytes.
600 =for apidoc Am|bool|memEQs|char* s1|STRLEN l1|"s2"
601 Like L</memEQ>, but the second string is a literal enclosed in double quotes,
602 C<l1> gives the number of bytes in C<s1>.
603 Returns true or false.
605 =for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
606 Test two buffers (which may contain embedded C<NUL> characters, to see if they
607 are not equal. The C<len> parameter indicates the number of bytes to compare.
608 Returns true or false. It is undefined behavior if either of the buffers
609 doesn't contain at least C<len> bytes.
611 =for apidoc Am|bool|memNEs|char* s1|STRLEN l1|"s2"
612 Like L</memNE>, but the second string is a literal enclosed in double quotes,
613 C<l1> gives the number of bytes in C<s1>.
614 Returns true or false.
616 =for apidoc Am|bool|memCHRs|"list"|char c
617 Returns the position of the first occurence of the byte C<c> in the literal
618 string C<"list">, or NULL if C<c> doesn't appear in C<"list">. All bytes are
619 treated as unsigned char. Thus this macro can be used to determine if C<c> is
620 in a set of particular characters. Unlike L<strchr(3)>, it works even if C<c>
621 is C<NUL> (and the set doesn't include C<NUL>).
625 New macros should use the following conventions for their names (which are
626 based on the underlying C library functions):
628 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
630 Each has two main parameters, string-like operands that are compared
631 against each other, as specified by the macro name. Some macros may
632 additionally have one or potentially even two length parameters. If a length
633 parameter applies to both string parameters, it will be positioned third;
634 otherwise any length parameter immediately follows the string parameter it
637 If the prefix to the name is 'str', the string parameter is a pointer to a C
638 language string. Such a string does not contain embedded NUL bytes; its
639 length may be unknown, but can be calculated by C<strlen()>, since it is
640 terminated by a NUL, which isn't included in its length.
642 The optional 'n' following 'str' means that there is a third parameter,
643 giving the maximum number of bytes to look at in each string. Even if both
644 strings are longer than the length parameter, those extra bytes will be
647 The 's' suffix means that the 2nd byte string parameter is a literal C
648 double-quoted string. Its length will automatically be calculated by the
649 macro, so no length parameter will ever be needed for it.
651 If the prefix is 'mem', the string parameters don't have to be C strings;
652 they may contain embedded NUL bytes, do not necessarily have a terminating
653 NUL, and their lengths can be known only through other means, which in
654 practice are additional parameter(s) passed to the function. All 'mem'
655 functions have at least one length parameter. Barring any 'l' or 's' suffix,
656 there is a single length parameter, in position 3, which applies to both
657 string parameters. The 's' suffix means, as described above, that the 2nd
658 string is a literal double-quoted C string (hence its length is calculated by
659 the macro, and the length parameter to the function applies just to the first
660 string parameter, and hence is positioned just after it). An 'l' suffix
661 means that the 2nd string parameter has its own length parameter, and the
662 signature will look like memFOOl(s1, l1, s2, l2).
664 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
665 substring of the 1st string. 'P' if present indicates that the substring
666 must be a "proper" one in tha mathematical sense that the first one must be
667 strictly larger than the 2nd.
672 #define strNE(s1,s2) (strcmp(s1,s2) != 0)
673 #define strEQ(s1,s2) (strcmp(s1,s2) == 0)
674 #define strLT(s1,s2) (strcmp(s1,s2) < 0)
675 #define strLE(s1,s2) (strcmp(s1,s2) <= 0)
676 #define strGT(s1,s2) (strcmp(s1,s2) > 0)
677 #define strGE(s1,s2) (strcmp(s1,s2) >= 0)
679 #define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
680 #define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
682 #define memEQ(s1,s2,l) (memcmp(((const void *) (s1)), ((const void *) (s2)), l) == 0)
683 #define memNE(s1,s2,l) (! memEQ(s1,s2,l))
685 /* memEQ and memNE where second comparand is a string constant */
686 #define memEQs(s1, l, s2) \
687 (((sizeof(s2)-1) == (l)) && memEQ((s1), ASSERT_IS_LITERAL(s2), (sizeof(s2)-1)))
688 #define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
690 /* Keep these private until we decide it was a good idea */
691 #if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
693 #define strBEGINs(s1,s2) (strncmp(s1,ASSERT_IS_LITERAL(s2), sizeof(s2)-1) == 0)
695 #define memBEGINs(s1, l, s2) \
696 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
697 && memEQ(s1, ASSERT_IS_LITERAL(s2), sizeof(s2)-1))
698 #define memBEGINPs(s1, l, s2) \
699 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) - 1 \
700 && memEQ(s1, ASSERT_IS_LITERAL(s2), sizeof(s2)-1))
701 #define memENDs(s1, l, s2) \
702 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
703 && memEQ(s1 + (l) - (sizeof(s2) - 1), ASSERT_IS_LITERAL(s2), sizeof(s2)-1))
704 #define memENDPs(s1, l, s2) \
705 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) \
706 && memEQ(s1 + (l) - (sizeof(s2) - 1), ASSERT_IS_LITERAL(s2), sizeof(s2)-1))
707 #endif /* End of making macros private */
709 #define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
710 #define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
711 #define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
712 #define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
714 #define memCHRs(s1,c) ((const char *) memchr(ASSERT_IS_LITERAL(s1) , c, sizeof(s1)-1))
719 * Unfortunately, the introduction of locales means that we
720 * can't trust isupper(), etc. to tell the truth. And when
721 * it comes to /\w+/ with tainting enabled, we *must* be able
722 * to trust our character classes.
724 * Therefore, the default tests in the text of Perl will be independent of
725 * locale. Any code that wants to depend on the current locale will use the
726 * macros that contain _LC in their names
729 #ifdef USE_LOCALE_CTYPE
737 =head1 Character classification
738 This section is about functions (really macros) that classify characters
739 into types, such as punctuation versus alphabetic, etc. Most of these are
740 analogous to regular expression character classes. (See
741 L<perlrecharclass/POSIX Character Classes>.) There are several variants for
742 each class. (Not all macros have all variants; each item below lists the
743 ones valid for it.) None are affected by C<use bytes>, and only the ones
744 with C<LC> in the name are affected by the current locale.
746 The base function, e.g., C<isALPHA()>, takes any signed or unsigned value,
747 treating it as a code point, and returns a boolean as to whether or not the
748 character represented by it is (or on non-ASCII platforms, corresponds to) an
749 ASCII character in the named class based on platform, Unicode, and Perl rules.
750 If the input is a number that doesn't fit in an octet, FALSE is returned.
752 Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
753 with no suffix C<"_A">. This variant is used to emphasize by its name that
754 only ASCII-range characters can return TRUE.
756 Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
757 onto the platform. That is, the code points that are ASCII are unaffected,
758 since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
759 as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
760 true when called with the code point 0xDF, which is a word character in both
761 ASCII and EBCDIC (though it represents different characters in each).
762 If the input is a number that doesn't fit in an octet, FALSE is returned.
763 (Perl's documentation uses a colloquial definition of Latin-1, to include all
764 code points below 256.)
766 Variant C<isI<FOO>_uvchr> is exactly like the C<isI<FOO>_L1> variant, for
767 inputs below 256, but if the code point is larger than 255, Unicode rules are
768 used to determine if it is in the character class. For example,
769 C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
770 WITH MACRON in Unicode, and is a word character.
772 Variants C<isI<FOO>_utf8> and C<isI<FOO>_utf8_safe> are like C<isI<FOO>_uvchr>,
773 but are used for UTF-8 encoded strings. The two forms are different names for
774 the same thing. Each call to one of these classifies the first character of
775 the string starting at C<p>. The second parameter, C<e>, points to anywhere in
776 the string beyond the first character, up to one byte past the end of the
777 entire string. Although both variants are identical, the suffix C<_safe> in
778 one name emphasizes that it will not attempt to read beyond S<C<e - 1>>,
779 provided that the constraint S<C<s E<lt> e>> is true (this is asserted for in
780 C<-DDEBUGGING> builds). If the UTF-8 for the input character is malformed in
781 some way, the program may croak, or the function may return FALSE, at the
782 discretion of the implementation, and subject to change in future releases.
784 Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants,
785 but the result is based on the current locale, which is what C<LC> in the name
786 stands for. If Perl can determine that the current locale is a UTF-8 locale,
787 it uses the published Unicode rules; otherwise, it uses the C library function
788 that gives the named classification. For example, C<isDIGIT_LC()> when not in
789 a UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
790 returned if the input won't fit into an octet. On some platforms where the C
791 library function is known to be defective, Perl changes its result to follow
792 the POSIX standard's rules.
794 Variant C<isI<FOO>_LC_uvchr> acts exactly like C<isI<FOO>_LC> for inputs less
795 than 256, but for larger ones it returns the Unicode classification of the code
798 Variants C<isI<FOO>_LC_utf8> and C<isI<FOO>_LC_utf8_safe> are like
799 C<isI<FOO>_LC_uvchr>, but are used for UTF-8 encoded strings. The two forms
800 are different names for the same thing. Each call to one of these classifies
801 the first character of the string starting at C<p>. The second parameter,
802 C<e>, points to anywhere in the string beyond the first character, up to one
803 byte past the end of the entire string. Although both variants are identical,
804 the suffix C<_safe> in one name emphasizes that it will not attempt to read
805 beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
806 is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
807 character is malformed in some way, the program may croak, or the function may
808 return FALSE, at the discretion of the implementation, and subject to change in
811 =for apidoc Am|bool|isALPHA|UV ch
812 =for apidoc_item ||isALPHA_A|UV ch
813 =for apidoc_item ||isALPHA_LC|UV ch
814 =for apidoc_item ||isALPHA_LC_utf8_safe|U8 * s| U8 *end
815 =for apidoc_item ||isALPHA_LC_uvchr|UV ch
816 =for apidoc_item ||isALPHA_L1|UV ch
817 =for apidoc_item ||isALPHA_utf8|U8 * s|U8 * end
818 =for apidoc_item ||isALPHA_utf8_safe|U8 * s|U8 * end
819 =for apidoc_item ||isALPHA_uvchr|UV ch
820 Returns a boolean indicating whether the specified input is one of C<[A-Za-z]>,
821 analogous to C<m/[[:alpha:]]/>.
822 See the L<top of this section|/Character classification> for an explanation of
827 Here and below, we add the prototypes of these macros for downstream programs
828 that would be interested in them, such as Devel::PPPort
830 =for apidoc Am|bool|isALPHANUMERIC|UV ch
831 =for apidoc_item ||isALPHANUMERIC_A|UV ch
832 =for apidoc_item ||isALPHANUMERIC_LC|UV ch
833 =for apidoc_item ||isALPHANUMERIC_LC_utf8_safe|U8 * s| U8 *end
834 =for apidoc_item ||isALPHANUMERIC_LC_uvchr|UV ch
835 =for apidoc_item ||isALPHANUMERIC_L1|UV ch
836 =for apidoc_item ||isALPHANUMERIC_utf8|U8 * s|U8 * end
837 =for apidoc_item ||isALPHANUMERIC_utf8_safe|U8 * s|U8 * end
838 =for apidoc_item ||isALPHANUMERIC_uvchr|UV ch
839 Returns a boolean indicating whether the specified character is one of
840 C<[A-Za-z0-9]>, analogous to C<m/[[:alnum:]]/>.
841 See the L<top of this section|/Character classification> for an explanation of
844 =for apidoc Am|bool|isALNUMC|UV ch
845 =for apidoc_item ||isALNUMC_A|UV ch
846 =for apidoc_item ||isALNUMC_LC|UV ch
847 =for apidoc_item ||isALNUMC_LC_uvchr|UV ch
848 =for apidoc_item ||isALNUMC_L1|UV ch
849 These are discouraged, backward compatibility macros for L</C<isALPHANUMERIC>>.
850 That is, each returns a boolean indicating whether the specified character is
851 one of C<[A-Za-z0-9]>, analogous to C<m/[[:alnum:]]/>.
853 The C<C> suffix in the names was meant to indicate that they correspond to the
854 C language L<C<isalnum(3)>>.
856 =for apidoc Am|bool|isASCII|UV ch
857 =for apidoc_item ||isASCII_A|UV ch
858 =for apidoc_item ||isASCII_LC|UV ch
859 =for apidoc_item ||isASCII_LC_utf8_safe|U8 * s| U8 *end
860 =for apidoc_item ||isASCII_LC_uvchr|UV ch
861 =for apidoc_item ||isASCII_L1|UV ch
862 =for apidoc_item ||isASCII_utf8|U8 * s|U8 * end
863 =for apidoc_item ||isASCII_utf8_safe|U8 * s|U8 * end
864 =for apidoc_item ||isASCII_uvchr|UV ch
865 Returns a boolean indicating whether the specified character is one of the 128
866 characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
867 On non-ASCII platforms, it returns TRUE iff this
868 character corresponds to an ASCII character. Variants C<isASCII_A()> and
869 C<isASCII_L1()> are identical to C<isASCII()>.
870 See the L<top of this section|/Character classification> for an explanation of
872 Note, however, that some platforms do not have the C library routine
873 C<isascii()>. In these cases, the variants whose names contain C<LC> are the
874 same as the corresponding ones without.
876 Also note, that because all ASCII characters are UTF-8 invariant (meaning they
877 have the exact same representation (always a single byte) whether encoded in
878 UTF-8 or not), C<isASCII> will give the correct results when called with any
879 byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8> and
880 C<isASCII_utf8_safe> will work properly on any string encoded or not in UTF-8.
882 =for apidoc Am|bool|isBLANK|UV ch
883 =for apidoc_item ||isBLANK_A|UV ch
884 =for apidoc_item ||isBLANK_LC|UV ch
885 =for apidoc_item ||isBLANK_LC_utf8_safe|U8 * s| U8 *end
886 =for apidoc_item ||isBLANK_LC_uvchr|UV ch
887 =for apidoc_item ||isBLANK_L1|UV ch
888 =for apidoc_item ||isBLANK_utf8|U8 * s|U8 * end
889 =for apidoc_item ||isBLANK_utf8_safe|U8 * s|U8 * end
890 =for apidoc_item ||isBLANK_uvchr|UV ch
891 Returns a boolean indicating whether the specified character is a
892 character considered to be a blank, analogous to C<m/[[:blank:]]/>.
893 See the L<top of this section|/Character classification> for an explanation of
896 however, that some platforms do not have the C library routine
897 C<isblank()>. In these cases, the variants whose names contain C<LC> are
898 the same as the corresponding ones without.
900 =for apidoc Am|bool|isCNTRL|UV ch
901 =for apidoc_item ||isCNTRL_A|UV ch
902 =for apidoc_item ||isCNTRL_LC|UV ch
903 =for apidoc_item ||isCNTRL_LC_utf8_safe|U8 * s| U8 *end
904 =for apidoc_item ||isCNTRL_LC_uvchr|UV ch
905 =for apidoc_item ||isCNTRL_L1|UV ch
906 =for apidoc_item ||isCNTRL_utf8|U8 * s|U8 * end
907 =for apidoc_item ||isCNTRL_utf8_safe|U8 * s|U8 * end
908 =for apidoc_item ||isCNTRL_uvchr|UV ch
910 Returns a boolean indicating whether the specified character is a
911 control character, analogous to C<m/[[:cntrl:]]/>.
912 See the L<top of this section|/Character classification> for an explanation of
914 On EBCDIC platforms, you almost always want to use the C<isCNTRL_L1> variant.
916 =for apidoc Am|bool|isDIGIT|UV ch
917 =for apidoc_item ||isDIGIT_A|UV ch
918 =for apidoc_item ||isDIGIT_LC|UV ch
919 =for apidoc_item ||isDIGIT_LC_utf8_safe|U8 * s| U8 *end
920 =for apidoc_item ||isDIGIT_LC_uvchr|UV ch
921 =for apidoc_item ||isDIGIT_L1|UV ch
922 =for apidoc_item ||isDIGIT_utf8|U8 * s|U8 * end
923 =for apidoc_item ||isDIGIT_utf8_safe|U8 * s|U8 * end
924 =for apidoc_item ||isDIGIT_uvchr|UV ch
926 Returns a boolean indicating whether the specified character is a
927 digit, analogous to C<m/[[:digit:]]/>.
928 Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
929 See the L<top of this section|/Character classification> for an explanation of
932 =for apidoc Am|bool|isGRAPH|UV ch
933 =for apidoc_item ||isGRAPH_A|UV ch
934 =for apidoc_item ||isGRAPH_LC|UV ch
935 =for apidoc_item ||isGRAPH_LC_utf8_safe|U8 * s| U8 *end
936 =for apidoc_item ||isGRAPH_LC_uvchr|UV ch
937 =for apidoc_item ||isGRAPH_L1|UV ch
938 =for apidoc_item ||isGRAPH_utf8|U8 * s|U8 * end
939 =for apidoc_item ||isGRAPH_utf8_safe|U8 * s|U8 * end
940 =for apidoc_item ||isGRAPH_uvchr|UV ch
941 Returns a boolean indicating whether the specified character is a
942 graphic character, analogous to C<m/[[:graph:]]/>.
943 See the L<top of this section|/Character classification> for an explanation of
946 =for apidoc Am|bool|isLOWER|UV ch
947 =for apidoc_item ||isLOWER_A|UV ch
948 =for apidoc_item ||isLOWER_LC|UV ch
949 =for apidoc_item ||isLOWER_LC_utf8_safe|U8 * s| U8 *end
950 =for apidoc_item ||isLOWER_LC_uvchr|UV ch
951 =for apidoc_item ||isLOWER_L1|UV ch
952 =for apidoc_item ||isLOWER_utf8|U8 * s|U8 * end
953 =for apidoc_item ||isLOWER_utf8_safe|U8 * s|U8 * end
954 =for apidoc_item ||isLOWER_uvchr|UV ch
955 Returns a boolean indicating whether the specified character is a
956 lowercase character, analogous to C<m/[[:lower:]]/>.
957 See the L<top of this section|/Character classification> for an explanation of
960 =for apidoc Am|bool|isOCTAL|UV ch
961 =for apidoc_item ||isOCTAL_A|UV ch
962 =for apidoc_item ||isOCTAL_L1|UV ch
963 Returns a boolean indicating whether the specified character is an
965 The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
968 =for apidoc Am|bool|isPUNCT|UV ch
969 =for apidoc_item ||isPUNCT_A|UV ch
970 =for apidoc_item ||isPUNCT_LC|UV ch
971 =for apidoc_item ||isPUNCT_LC_utf8_safe|U8 * s| U8 *end
972 =for apidoc_item ||isPUNCT_LC_uvchr|UV ch
973 =for apidoc_item ||isPUNCT_L1|UV ch
974 =for apidoc_item ||isPUNCT_utf8|U8 * s|U8 * end
975 =for apidoc_item ||isPUNCT_utf8_safe|U8 * s|U8 * end
976 =for apidoc_item ||isPUNCT_uvchr|UV ch
977 Returns a boolean indicating whether the specified character is a
978 punctuation character, analogous to C<m/[[:punct:]]/>.
979 Note that the definition of what is punctuation isn't as
980 straightforward as one might desire. See L<perlrecharclass/POSIX Character
981 Classes> for details.
982 See the L<top of this section|/Character classification> for an explanation of
985 =for apidoc Am|bool|isSPACE|UV ch
986 =for apidoc_item ||isSPACE_A|UV ch
987 =for apidoc_item ||isSPACE_LC|UV ch
988 =for apidoc_item ||isSPACE_LC_utf8_safe|U8 * s| U8 *end
989 =for apidoc_item ||isSPACE_LC_uvchr|UV ch
990 =for apidoc_item ||isSPACE_L1|UV ch
991 =for apidoc_item ||isSPACE_utf8|U8 * s|U8 * end
992 =for apidoc_item ||isSPACE_utf8_safe|U8 * s|U8 * end
993 =for apidoc_item ||isSPACE_uvchr|UV ch
994 Returns a boolean indicating whether the specified character is a
995 whitespace character. This is analogous
996 to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
997 this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
998 locale forms of this macro (the ones with C<LC> in their names) matched
999 precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
1000 in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
1001 (See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
1002 See the L<top of this section|/Character classification> for an explanation of
1005 =for apidoc Am|bool|isPSXSPC|UV ch
1006 =for apidoc_item ||isPSXSPC_A|UV ch
1007 =for apidoc_item ||isPSXSPC_LC|UV ch
1008 =for apidoc_item ||isPSXSPC_LC_utf8_safe|U8 * s| U8 *end
1009 =for apidoc_item ||isPSXSPC_LC_uvchr|UV ch
1010 =for apidoc_item ||isPSXSPC_L1|UV ch
1011 =for apidoc_item ||isPSXSPC_utf8|U8 * s|U8 * end
1012 =for apidoc_item ||isPSXSPC_utf8_safe|U8 * s|U8 * end
1013 =for apidoc_item ||isPSXSPC_uvchr|UV ch
1014 (short for Posix Space)
1015 Starting in 5.18, this is identical in all its forms to the
1016 corresponding C<isSPACE()> macros.
1017 The locale forms of this macro are identical to their corresponding
1018 C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
1019 non-locale forms differ from their C<isSPACE()> forms only in that the
1020 C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
1021 Otherwise they are identical. Thus this macro is analogous to what
1022 C<m/[[:space:]]/> matches in a regular expression.
1023 See the L<top of this section|/Character classification> for an explanation of
1026 =for apidoc Am|bool|isUPPER|UV ch
1027 =for apidoc_item ||isUPPER_A|UV ch
1028 =for apidoc_item ||isUPPER_LC|UV ch
1029 =for apidoc_item ||isUPPER_LC_utf8_safe|U8 * s| U8 *end
1030 =for apidoc_item ||isUPPER_LC_uvchr|UV ch
1031 =for apidoc_item ||isUPPER_L1|UV ch
1032 =for apidoc_item ||isUPPER_utf8|U8 * s|U8 * end
1033 =for apidoc_item ||isUPPER_utf8_safe|U8 * s|U8 * end
1034 =for apidoc_item ||isUPPER_uvchr|UV ch
1035 Returns a boolean indicating whether the specified character is an
1036 uppercase character, analogous to C<m/[[:upper:]]/>.
1037 See the L<top of this section|/Character classification> for an explanation of
1040 =for apidoc Am|bool|isPRINT|UV ch
1041 =for apidoc_item ||isPRINT_A|UV ch
1042 =for apidoc_item ||isPRINT_LC|UV ch
1043 =for apidoc_item ||isPRINT_LC_utf8_safe|U8 * s| U8 *end
1044 =for apidoc_item ||isPRINT_LC_uvchr|UV ch
1045 =for apidoc_item ||isPRINT_L1|UV ch
1046 =for apidoc_item ||isPRINT_utf8|U8 * s|U8 * end
1047 =for apidoc_item ||isPRINT_utf8_safe|U8 * s|U8 * end
1048 =for apidoc_item ||isPRINT_uvchr|UV ch
1049 Returns a boolean indicating whether the specified character is a
1050 printable character, analogous to C<m/[[:print:]]/>.
1051 See the L<top of this section|/Character classification> for an explanation of
1054 =for apidoc Am|bool|isWORDCHAR|UV ch
1055 =for apidoc_item ||isWORDCHAR_A|UV ch
1056 =for apidoc_item ||isWORDCHAR_LC|UV ch
1057 =for apidoc_item ||isWORDCHAR_LC_utf8_safe|U8 * s| U8 *end
1058 =for apidoc_item ||isWORDCHAR_LC_uvchr|UV ch
1059 =for apidoc_item ||isWORDCHAR_L1|UV ch
1060 =for apidoc_item ||isWORDCHAR_utf8|U8 * s|U8 * end
1061 =for apidoc_item ||isWORDCHAR_utf8_safe|U8 * s|U8 * end
1062 =for apidoc_item ||isWORDCHAR_uvchr|UV ch
1063 Returns a boolean indicating whether the specified character is a character
1064 that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
1065 in a regular expression. A word character is an alphabetic character, a
1066 decimal digit, a connecting punctuation character (such as an underscore), or
1067 a "mark" character that attaches to one of those (like some sort of accent).
1069 See the L<top of this section|/Character classification> for an explanation of
1072 C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>,
1073 C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, C<isWORDCHAR_LC_utf8>, and
1074 C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
1075 include the platform's native underscore.
1077 =for apidoc Am|bool|isALNUM |UV ch
1078 =for apidoc_item ||isALNUM_A |UV ch
1079 =for apidoc_item ||isALNUM_LC |UV ch
1080 =for apidoc_item ||isALNUM_LC_uvchr|UV ch
1081 These are each a synonym for their respectively named L</C<isWORDCHAR>>
1084 They are provided for backward compatibility, even though a word character
1085 includes more than the standard C language meaning of alphanumeric.
1086 To get the C language definition, use the corresponding L</C<isALPHANUMERIC>>
1089 =for apidoc Am|bool|isXDIGIT|UV ch
1090 =for apidoc_item ||isXDIGIT_A|UV ch
1091 =for apidoc_item ||isXDIGIT_LC|UV ch
1092 =for apidoc_item ||isXDIGIT_LC_utf8_safe|U8 * s| U8 *end
1093 =for apidoc_item ||isXDIGIT_LC_uvchr|UV ch
1094 =for apidoc_item ||isXDIGIT_L1|UV ch
1095 =for apidoc_item ||isXDIGIT_utf8|U8 * s|U8 * end
1096 =for apidoc_item ||isXDIGIT_utf8_safe|U8 * s|U8 * end
1097 =for apidoc_item ||isXDIGIT_uvchr|UV ch
1098 Returns a boolean indicating whether the specified character is a hexadecimal
1099 digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
1100 and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
1101 See the L<top of this section|/Character classification> for an explanation of
1104 =for apidoc Am|bool|isIDFIRST|UV ch
1105 =for apidoc_item ||isIDFIRST_A|UV ch
1106 =for apidoc_item ||isIDFIRST_LC|UV ch
1107 =for apidoc_item ||isIDFIRST_LC_utf8_safe|U8 * s| U8 *end
1108 =for apidoc_item ||isIDFIRST_LC_uvchr|UV ch
1109 =for apidoc_item ||isIDFIRST_L1|UV ch
1110 =for apidoc_item ||isIDFIRST_utf8|U8 * s|U8 * end
1111 =for apidoc_item ||isIDFIRST_utf8_safe|U8 * s|U8 * end
1112 =for apidoc_item ||isIDFIRST_uvchr|UV ch
1113 Returns a boolean indicating whether the specified character can be the first
1114 character of an identifier. This is very close to, but not quite the same as
1115 the official Unicode property C<XID_Start>. The difference is that this
1116 returns true only if the input character also matches L</isWORDCHAR>.
1117 See the L<top of this section|/Character classification> for an explanation of
1120 =for apidoc Am|bool|isIDCONT|UV ch
1121 =for apidoc_item ||isIDCONT_A|UV ch
1122 =for apidoc_item ||isIDCONT_LC|UV ch
1123 =for apidoc_item ||isIDCONT_LC_utf8_safe|U8 * s| U8 *end
1124 =for apidoc_item ||isIDCONT_LC_uvchr|UV ch
1125 =for apidoc_item ||isIDCONT_L1|UV ch
1126 =for apidoc_item ||isIDCONT_utf8|U8 * s|U8 * end
1127 =for apidoc_item ||isIDCONT_utf8_safe|U8 * s|U8 * end
1128 =for apidoc_item ||isIDCONT_uvchr|UV ch
1129 Returns a boolean indicating whether the specified character can be the
1130 second or succeeding character of an identifier. This is very close to, but
1131 not quite the same as the official Unicode property C<XID_Continue>. The
1132 difference is that this returns true only if the input character also matches
1133 L</isWORDCHAR>. See the L<top of this section|/Character classification> for
1134 an explanation of the variants.
1136 =for apidoc_section $numeric
1138 =for apidoc Am|U8|READ_XDIGIT|char str*
1139 Returns the value of an ASCII-range hex digit and advances the string pointer.
1140 Behaviour is only well defined when isXDIGIT(*str) is true.
1142 =head1 Character case changing
1143 Perl uses "full" Unicode case mappings. This means that converting a single
1144 character to another case may result in a sequence of more than one character.
1145 For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
1146 character sequence C<SS>. This presents some complications The lowercase of
1147 all characters in the range 0..255 is a single character, and thus
1148 C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
1149 return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
1150 an API that does allow every possible legal result to be returned.) Likewise
1151 no other function that is crippled by not being able to give the correct
1152 results for the full range of possible inputs has been implemented here.
1154 =for apidoc Am|UV|toUPPER|UV cp
1155 =for apidoc_item |UV|toUPPER_A|UV cp
1156 =for apidoc_item |UV|toUPPER_utf8|U8* p|U8* e|U8* s|STRLEN* lenp
1157 =for apidoc_item |UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1158 =for apidoc_item |UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
1160 These all return the uppercase of a character. The differences are what domain
1161 they operate on, and whether the input is specified as a code point (those
1162 forms with a C<cp> parameter) or as a UTF-8 string (the others). In the latter
1163 case, the code point to use is the first one in the buffer of UTF-8 encoded
1164 code points, delineated by the arguments S<C<p .. e - 1>>.
1166 C<toUPPER> and C<toUPPER_A> are synonyms of each other. They return the
1167 uppercase of any lowercase ASCII-range code point. All other inputs are
1168 returned unchanged. Since these are macros, the input type may be any integral
1169 one, and the output will occupy the same number of bits as the input.
1171 There is no C<toUPPER_L1> nor C<toUPPER_LATIN1> as the uppercase of some code
1172 points in the 0..255 range is above that range or consists of multiple
1173 characters. Instead use C<toUPPER_uvchr>.
1175 C<toUPPER_uvchr> returns the uppercase of any Unicode code point. The return
1176 value is identical to that of C<toUPPER_A> for input code points in the ASCII
1177 range. The uppercase of the vast majority of Unicode code points is the same
1178 as the code point itself. For these, and for code points above the legal
1179 Unicode maximum, this returns the input code point unchanged. It additionally
1180 stores the UTF-8 of the result into the buffer beginning at C<s>, and its
1181 length in bytes into C<*lenp>. The caller must have made C<s> large enough to
1182 contain at least C<UTF8_MAXBYTES_CASE+1> bytes to avoid possible overflow.
1184 NOTE: the uppercase of a code point may be more than one code point. The
1185 return value of this function is only the first of these. The entire uppercase
1186 is returned in C<s>. To determine if the result is more than a single code
1187 point, you can do something like this:
1189 uc = toUPPER_uvchr(cp, s, &len);
1190 if (len > UTF8SKIP(s)) { is multiple code points }
1191 else { is a single code point }
1193 C<toUPPER_utf8> and C<toUPPER_utf8_safe> are synonyms of each other. The only
1194 difference between these and C<toUPPER_uvchr> is that the source for these is
1195 encoded in UTF-8, instead of being a code point. It is passed as a buffer
1196 starting at C<p>, with C<e> pointing to one byte beyond its end. The C<p>
1197 buffer may certainly contain more than one code point; but only the first one
1198 (up through S<C<e - 1>>) is examined. If the UTF-8 for the input character is
1199 malformed in some way, the program may croak, or the function may return the
1200 REPLACEMENT CHARACTER, at the discretion of the implementation, and subject to
1201 change in future releases.
1203 =for apidoc Am|UV|toFOLD|UV cp
1204 =for apidoc_item |UV|toFOLD_A|UV cp
1205 =for apidoc_item |UV|toFOLD_utf8|U8* p|U8* e|U8* s|STRLEN* lenp
1206 =for apidoc_item |UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1207 =for apidoc_item |UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
1209 These all return the foldcase of a character. "foldcase" is an internal case
1210 for C</i> pattern matching. If the foldcase of character A and the foldcase of
1211 character B are the same, they match caselessly; otherwise they don't.
1213 The differences in the forms are what domain they operate on, and whether the
1214 input is specified as a code point (those forms with a C<cp> parameter) or as a
1215 UTF-8 string (the others). In the latter case, the code point to use is the
1216 first one in the buffer of UTF-8 encoded code points, delineated by the
1217 arguments S<C<p .. e - 1>>.
1219 C<toFOLD> and C<toFOLD_A> are synonyms of each other. They return the
1220 foldcase of any ASCII-range code point. In this range, the foldcase is
1221 identical to the lowercase. All other inputs are returned unchanged. Since
1222 these are macros, the input type may be any integral one, and the output will
1223 occupy the same number of bits as the input.
1225 There is no C<toFOLD_L1> nor C<toFOLD_LATIN1> as the foldcase of some code
1226 points in the 0..255 range is above that range or consists of multiple
1227 characters. Instead use C<toFOLD_uvchr>.
1229 C<toFOLD_uvchr> returns the foldcase of any Unicode code point. The return
1230 value is identical to that of C<toFOLD_A> for input code points in the ASCII
1231 range. The foldcase of the vast majority of Unicode code points is the same
1232 as the code point itself. For these, and for code points above the legal
1233 Unicode maximum, this returns the input code point unchanged. It additionally
1234 stores the UTF-8 of the result into the buffer beginning at C<s>, and its
1235 length in bytes into C<*lenp>. The caller must have made C<s> large enough to
1236 contain at least C<UTF8_MAXBYTES_CASE+1> bytes to avoid possible overflow.
1238 NOTE: the foldcase of a code point may be more than one code point. The
1239 return value of this function is only the first of these. The entire foldcase
1240 is returned in C<s>. To determine if the result is more than a single code
1241 point, you can do something like this:
1243 uc = toFOLD_uvchr(cp, s, &len);
1244 if (len > UTF8SKIP(s)) { is multiple code points }
1245 else { is a single code point }
1247 C<toFOLD_utf8> and C<toFOLD_utf8_safe> are synonyms of each other. The only
1248 difference between these and C<toFOLD_uvchr> is that the source for these is
1249 encoded in UTF-8, instead of being a code point. It is passed as a buffer
1250 starting at C<p>, with C<e> pointing to one byte beyond its end. The C<p>
1251 buffer may certainly contain more than one code point; but only the first one
1252 (up through S<C<e - 1>>) is examined. If the UTF-8 for the input character is
1253 malformed in some way, the program may croak, or the function may return the
1254 REPLACEMENT CHARACTER, at the discretion of the implementation, and subject to
1255 change in future releases.
1257 =for apidoc Am|UV|toLOWER|UV cp
1258 =for apidoc_item |UV|toLOWER_A|UV cp
1259 =for apidoc_item |UV|toLOWER_LATIN1|UV cp
1260 =for apidoc_item |UV|toLOWER_LC|UV cp
1261 =for apidoc_item |UV|toLOWER_L1|UV cp
1262 =for apidoc_item |UV|toLOWER_utf8|U8* p|U8* e|U8* s|STRLEN* lenp
1263 =for apidoc_item |UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1264 =for apidoc_item |UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
1266 These all return the lowercase of a character. The differences are what domain
1267 they operate on, and whether the input is specified as a code point (those
1268 forms with a C<cp> parameter) or as a UTF-8 string (the others). In the latter
1269 case, the code point to use is the first one in the buffer of UTF-8 encoded
1270 code points, delineated by the arguments S<C<p .. e - 1>>.
1272 C<toLOWER> and C<toLOWER_A> are synonyms of each other. They return the
1273 lowercase of any uppercase ASCII-range code point. All other inputs are
1274 returned unchanged. Since these are macros, the input type may be any integral
1275 one, and the output will occupy the same number of bits as the input.
1277 C<toLOWER_L1> and C<toLOWER_LATIN1> are synonyms of each other. They behave
1278 identically as C<toLOWER> for ASCII-range input. But additionally will return
1279 the lowercase of any uppercase code point in the entire 0..255 range, assuming
1280 a Latin-1 encoding (or the EBCDIC equivalent on such platforms).
1282 C<toLOWER_LC> returns the lowercase of the input code point according to the
1283 rules of the current POSIX locale. Input code points outside the range 0..255
1284 are returned unchanged.
1286 C<toLOWER_uvchr> returns the lowercase of any Unicode code point. The return
1287 value is identical to that of C<toLOWER_L1> for input code points in the 0..255
1288 range. The lowercase of the vast majority of Unicode code points is the same
1289 as the code point itself. For these, and for code points above the legal
1290 Unicode maximum, this returns the input code point unchanged. It additionally
1291 stores the UTF-8 of the result into the buffer beginning at C<s>, and its
1292 length in bytes into C<*lenp>. The caller must have made C<s> large enough to
1293 contain at least C<UTF8_MAXBYTES_CASE+1> bytes to avoid possible overflow.
1295 NOTE: the lowercase of a code point may be more than one code point. The
1296 return value of this function is only the first of these. The entire lowercase
1297 is returned in C<s>. To determine if the result is more than a single code
1298 point, you can do something like this:
1300 uc = toLOWER_uvchr(cp, s, &len);
1301 if (len > UTF8SKIP(s)) { is multiple code points }
1302 else { is a single code point }
1304 C<toLOWER_utf8> and C<toLOWER_utf8_safe> are synonyms of each other. The only
1305 difference between these and C<toLOWER_uvchr> is that the source for these is
1306 encoded in UTF-8, instead of being a code point. It is passed as a buffer
1307 starting at C<p>, with C<e> pointing to one byte beyond its end. The C<p>
1308 buffer may certainly contain more than one code point; but only the first one
1309 (up through S<C<e - 1>>) is examined. If the UTF-8 for the input character is
1310 malformed in some way, the program may croak, or the function may return the
1311 REPLACEMENT CHARACTER, at the discretion of the implementation, and subject to
1312 change in future releases.
1314 =for apidoc Am|UV|toTITLE|UV cp
1315 =for apidoc_item |UV|toTITLE_A|UV cp
1316 =for apidoc_item |UV|toTITLE_utf8|U8* p|U8* e|U8* s|STRLEN* lenp
1317 =for apidoc_item |UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1318 =for apidoc_item |UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
1320 These all return the titlecase of a character. The differences are what domain
1321 they operate on, and whether the input is specified as a code point (those
1322 forms with a C<cp> parameter) or as a UTF-8 string (the others). In the latter
1323 case, the code point to use is the first one in the buffer of UTF-8 encoded
1324 code points, delineated by the arguments S<C<p .. e - 1>>.
1326 C<toTITLE> and C<toTITLE_A> are synonyms of each other. They return the
1327 titlecase of any lowercase ASCII-range code point. In this range, the
1328 titlecase is identical to the uppercase. All other inputs are returned
1329 unchanged. Since these are macros, the input type may be any integral one, and
1330 the output will occupy the same number of bits as the input.
1332 There is no C<toTITLE_L1> nor C<toTITLE_LATIN1> as the titlecase of some code
1333 points in the 0..255 range is above that range or consists of multiple
1334 characters. Instead use C<toTITLE_uvchr>.
1336 C<toTITLE_uvchr> returns the titlecase of any Unicode code point. The return
1337 value is identical to that of C<toTITLE_A> for input code points in the ASCII
1338 range. The titlecase of the vast majority of Unicode code points is the same
1339 as the code point itself. For these, and for code points above the legal
1340 Unicode maximum, this returns the input code point unchanged. It additionally
1341 stores the UTF-8 of the result into the buffer beginning at C<s>, and its
1342 length in bytes into C<*lenp>. The caller must have made C<s> large enough to
1343 contain at least C<UTF8_MAXBYTES_CASE+1> bytes to avoid possible overflow.
1345 NOTE: the titlecase of a code point may be more than one code point. The
1346 return value of this function is only the first of these. The entire titlecase
1347 is returned in C<s>. To determine if the result is more than a single code
1348 point, you can do something like this:
1350 uc = toTITLE_uvchr(cp, s, &len);
1351 if (len > UTF8SKIP(s)) { is multiple code points }
1352 else { is a single code point }
1354 C<toTITLE_utf8> and C<toTITLE_utf8_safe> are synonyms of each other. The only
1355 difference between these and C<toTITLE_uvchr> is that the source for these is
1356 encoded in UTF-8, instead of being a code point. It is passed as a buffer
1357 starting at C<p>, with C<e> pointing to one byte beyond its end. The C<p>
1358 buffer may certainly contain more than one code point; but only the first one
1359 (up through S<C<e - 1>>) is examined. If the UTF-8 for the input character is
1360 malformed in some way, the program may croak, or the function may return the
1361 REPLACEMENT CHARACTER, at the discretion of the implementation, and subject to
1362 change in future releases.
1366 XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1367 really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1368 and aren't general purpose as they don't work on U+DF, and assert against that.
1369 and isCASED_LC, as it really is more of an internal thing.
1371 Note that these macros are repeated in Devel::PPPort, so should also be
1372 patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1377 void below because that's the best fit, and works for Devel::PPPort
1378 =for apidoc_section $integer
1379 =for apidoc AyT||WIDEST_UTYPE
1381 Yields the widest unsigned integer type on the platform, currently either
1382 C<U32> or C<U64>. This can be used in declarations such as
1388 my_uv = (WIDEST_UTYPE) val;
1393 #define WIDEST_UTYPE PERL_UINTMAX_T
1395 /* Where there could be some confusion, use this as a static assert in macros
1396 * to make sure that a parameter isn't a pointer. But some compilers can't
1397 * handle this. The only one known so far that doesn't is gcc 3.3.6; the check
1398 * below isn't thorough for such an old compiler, so may have to be revised if
1399 * experience so dictates. */
1400 #if ! PERL_IS_GCC || PERL_GCC_VERSION_GT(3,3,6)
1401 # define ASSERT_NOT_PTR(x) ((x) | 0)
1403 # define ASSERT_NOT_PTR(x) (x)
1406 /* Likewise, this is effectively a static assert to be used to guarantee the
1407 * parameter is a pointer
1409 * NOT suitable for void*
1411 #define ASSERT_IS_PTR(x) (__ASSERT_(sizeof(*(x))) (x))
1413 /* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1414 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1415 * bits of information are lost even on a 64-bit machine, but to get the
1416 * compiler to optimize it out if possible. This is because Configure makes
1417 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1418 * sizeof() guarantees that this evaluates to a constant true at compile time.
1420 * For Coverity, be always true, because otherwise Coverity thinks
1421 * it finds several expressions that are always true, independent
1422 * of operands. Well, they are, but that is kind of the point.
1424 #ifndef __COVERITY__
1425 /* The '| 0' part in ASSERT_NOT_PTR ensures a compiler error if c is not
1426 * integer (like e.g., a pointer) */
1427 # define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
1428 || (((WIDEST_UTYPE) ASSERT_NOT_PTR(c)) >> 8) == 0)
1430 # define FITS_IN_8_BITS(c) (1)
1433 /* Returns true if l <= c <= (l + n), where 'l' and 'n' are non-negative
1434 * Written this way so that after optimization, only one conditional test is
1435 * needed. (The NV casts stop any warnings about comparison always being true
1436 * if called with an unsigned. The cast preserves the sign, which is all we
1438 #define withinCOUNT(c, l, n) (__ASSERT_((NV) (l) >= 0) \
1439 __ASSERT_((NV) (n) >= 0) \
1440 withinCOUNT_KNOWN_VALID_((c), (l), (n)))
1442 /* For internal use only, this can be used in places where it is known that the
1443 * parameters to withinCOUNT() are valid, to avoid the asserts. For example,
1444 * inRANGE() below, calls this several times, but does all the necessary
1445 * asserts itself, once. The reason that this is necessary is that the
1446 * duplicate asserts were exceeding the internal limits of some compilers */
1447 #define withinCOUNT_KNOWN_VALID_(c, l, n) \
1448 ((((WIDEST_UTYPE) (c)) - ASSERT_NOT_PTR(l)) \
1449 <= ((WIDEST_UTYPE) ASSERT_NOT_PTR(n)))
1451 /* Returns true if c is in the range l..u, where 'l' is non-negative
1452 * Written this way so that after optimization, only one conditional test is
1454 #define inRANGE(c, l, u) (__ASSERT_((NV) (l) >= 0) __ASSERT_((u) >= (l)) \
1455 ( (sizeof(c) == sizeof(U8)) ? inRANGE_helper_(U8, (c), (l), ((u))) \
1456 : (sizeof(c) == sizeof(U16)) ? inRANGE_helper_(U16,(c), (l), ((u))) \
1457 : (sizeof(c) == sizeof(U32)) ? inRANGE_helper_(U32,(c), (l), ((u))) \
1458 : (__ASSERT_(sizeof(c) == sizeof(WIDEST_UTYPE)) \
1459 inRANGE_helper_(WIDEST_UTYPE,(c), (l), ((u))))))
1461 /* For internal use, this is used by machine-generated code which generates
1462 * known valid calls, with a known sizeof(). This avoids the extra code and
1463 * asserts that were exceeding internal limits of some compilers. */
1464 #define inRANGE_helper_(cast, c, l, u) \
1465 withinCOUNT_KNOWN_VALID_(((cast) (c)), (l), ((u) - (l)))
1468 # ifndef _ALL_SOURCE
1469 /* The native libc isascii() et.al. functions return the wrong results
1470 * on at least z/OS unless this is defined. */
1471 # error _ALL_SOURCE should probably be defined
1474 /* There is a simple definition of ASCII for ASCII platforms. But the
1475 * EBCDIC one isn't so simple, so is defined using table look-up like the
1476 * other macros below.
1478 * The cast here is used instead of '(c) >= 0', because some compilers emit
1479 * a warning that that test is always true when the parameter is an
1480 * unsigned type. khw supposes that it could be written as
1481 * && ((c) == '\0' || (c) > 0)
1482 * to avoid the message, but the cast will likely avoid extra branches even
1483 * with stupid compilers. */
1484 # define isASCII(c) (((WIDEST_UTYPE) ASSERT_NOT_PTR(c)) < 128)
1487 /* Take the eight possible bit patterns of the lower 3 bits and you get the
1488 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1489 * can be ignored. If the rest match '0', we have an octal */
1490 #define isOCTAL_A(c) ((((WIDEST_UTYPE) ASSERT_NOT_PTR(c)) & ~7) == '0')
1492 #ifdef H_PERL /* If have access to perl.h, lookup in its table */
1494 /* Character class numbers. For internal core Perl use only. The ones less
1495 * than 32 are used in PL_charclass[] and the ones up through the one that
1496 * corresponds to <HIGHEST_REGCOMP_DOT_H_SYNC_> are used by regcomp.h and
1497 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1498 * their actual definitions are here. If that file has a name not used here,
1501 * The first group of these is ordered in what I (khw) estimate to be the
1502 * frequency of their use. This gives a slight edge to exiting a loop earlier
1503 * (in reginclass() in regexec.c). Except \v should be last, as it isn't a
1504 * real Posix character class, and some (small) inefficiencies in regular
1505 * expression handling would be introduced by putting it in the middle of those
1506 * that are. Also, cntrl and ascii come after the others as it may be useful
1507 * to group these which have no members that match above Latin1, (or above
1508 * ASCII in the latter case) */
1510 # define CC_WORDCHAR_ 0 /* \w and [:word:] */
1511 # define CC_DIGIT_ 1 /* \d and [:digit:] */
1512 # define CC_ALPHA_ 2 /* [:alpha:] */
1513 # define CC_LOWER_ 3 /* [:lower:] */
1514 # define CC_UPPER_ 4 /* [:upper:] */
1515 # define CC_PUNCT_ 5 /* [:punct:] */
1516 # define CC_PRINT_ 6 /* [:print:] */
1517 # define CC_ALPHANUMERIC_ 7 /* [:alnum:] */
1518 # define CC_GRAPH_ 8 /* [:graph:] */
1519 # define CC_CASED_ 9 /* [:lower:] or [:upper:] under /i */
1520 # define CC_SPACE_ 10 /* \s, [:space:] */
1521 # define CC_BLANK_ 11 /* [:blank:] */
1522 # define CC_XDIGIT_ 12 /* [:xdigit:] */
1523 # define CC_CNTRL_ 13 /* [:cntrl:] */
1524 # define CC_ASCII_ 14 /* [:ascii:] */
1525 # define CC_VERTSPACE_ 15 /* \v */
1527 # define HIGHEST_REGCOMP_DOT_H_SYNC_ CC_VERTSPACE_
1529 /* The members of the third group below do not need to be coordinated with data
1530 * structures in regcomp.[ch] and regexec.c. */
1531 # define CC_IDFIRST_ 16
1532 # define CC_CHARNAME_CONT_ 17
1533 # define CC_NONLATIN1_FOLD_ 18
1534 # define CC_NONLATIN1_SIMPLE_FOLD_ 19
1535 # define CC_QUOTEMETA_ 20
1536 # define CC_NON_FINAL_FOLD_ 21
1537 # define CC_IS_IN_SOME_FOLD_ 22
1538 # define CC_BINDIGIT_ 23
1539 # define CC_OCTDIGIT_ 24
1540 # define CC_MNEMONIC_CNTRL_ 25
1543 * If more bits are needed, one could add a second word for non-64bit
1544 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
1545 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1546 * is used only for optimization (as of this writing), and differs in the
1547 * Latin1 range from the ALPHA bit only in two relatively unimportant
1548 * characters: the masculine and feminine ordinal indicators, so removing it
1549 * would just cause /i regexes which match them to run less efficiently.
1550 * Similarly the EBCDIC-only bits are used just for speed, and could be
1551 * replaced by other means */
1553 #if defined(PERL_CORE) || defined(PERL_EXT)
1554 /* An enum version of the character class numbers, to help compilers
1557 CC_ENUM_ALPHA_ = CC_ALPHA_,
1558 CC_ENUM_ALPHANUMERIC_ = CC_ALPHANUMERIC_,
1559 CC_ENUM_ASCII_ = CC_ASCII_,
1560 CC_ENUM_BLANK_ = CC_BLANK_,
1561 CC_ENUM_CASED_ = CC_CASED_,
1562 CC_ENUM_CNTRL_ = CC_CNTRL_,
1563 CC_ENUM_DIGIT_ = CC_DIGIT_,
1564 CC_ENUM_GRAPH_ = CC_GRAPH_,
1565 CC_ENUM_LOWER_ = CC_LOWER_,
1566 CC_ENUM_PRINT_ = CC_PRINT_,
1567 CC_ENUM_PUNCT_ = CC_PUNCT_,
1568 CC_ENUM_SPACE_ = CC_SPACE_,
1569 CC_ENUM_UPPER_ = CC_UPPER_,
1570 CC_ENUM_VERTSPACE_ = CC_VERTSPACE_,
1571 CC_ENUM_WORDCHAR_ = CC_WORDCHAR_,
1572 CC_ENUM_XDIGIT_ = CC_XDIGIT_
1573 } char_class_number_;
1576 #define POSIX_CC_COUNT (HIGHEST_REGCOMP_DOT_H_SYNC_ + 1)
1580 EXTCONST U32 PL_charclass[] = {
1581 # include "l1_char_class_tab.h"
1584 # else /* ! DOINIT */
1585 EXTCONST U32 PL_charclass[];
1589 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
1590 # define CC_mask_(classnum) (1U << (classnum))
1592 /* For internal core Perl use only: the base macro for defining macros like
1594 # define generic_isCC_(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
1595 && (PL_charclass[(U8) (c)] & CC_mask_(classnum)))
1597 /* The mask for the _A versions of the macros; it just adds in the bit for
1599 # define CC_mask_A_(classnum) (CC_mask_(classnum) | CC_mask_(CC_ASCII_))
1601 /* For internal core Perl use only: the base macro for defining macros like
1602 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1603 * the ASCII bit are present */
1604 # define generic_isCC_A_(c, classnum) (FITS_IN_8_BITS(c) \
1605 && ((PL_charclass[(U8) (c)] & CC_mask_A_(classnum)) \
1606 == CC_mask_A_(classnum)))
1608 /* On ASCII platforms certain classes form a single range. It's faster to
1609 * special case these. isDIGIT is a single range on all platforms */
1611 # define isALPHA_A(c) generic_isCC_A_(c, CC_ALPHA_)
1612 # define isGRAPH_A(c) generic_isCC_A_(c, CC_GRAPH_)
1613 # define isLOWER_A(c) generic_isCC_A_(c, CC_LOWER_)
1614 # define isPRINT_A(c) generic_isCC_A_(c, CC_PRINT_)
1615 # define isUPPER_A(c) generic_isCC_A_(c, CC_UPPER_)
1617 /* By folding the upper and lowercase, we can use a single range */
1618 # define isALPHA_A(c) inRANGE((~('A' ^ 'a') & (c)), 'A', 'Z')
1619 # define isGRAPH_A(c) inRANGE(c, ' ' + 1, 0x7e)
1620 # define isLOWER_A(c) inRANGE(c, 'a', 'z')
1621 # define isPRINT_A(c) inRANGE(c, ' ', 0x7e)
1622 # define isUPPER_A(c) inRANGE(c, 'A', 'Z')
1624 # define isALPHANUMERIC_A(c) generic_isCC_A_(c, CC_ALPHANUMERIC_)
1625 # define isBLANK_A(c) generic_isCC_A_(c, CC_BLANK_)
1626 # define isCNTRL_A(c) generic_isCC_A_(c, CC_CNTRL_)
1627 # define isDIGIT_A(c) inRANGE(c, '0', '9')
1628 # define isPUNCT_A(c) generic_isCC_A_(c, CC_PUNCT_)
1629 # define isSPACE_A(c) generic_isCC_A_(c, CC_SPACE_)
1630 # define isWORDCHAR_A(c) generic_isCC_A_(c, CC_WORDCHAR_)
1631 # define isXDIGIT_A(c) generic_isCC_(c, CC_XDIGIT_) /* No non-ASCII xdigits
1633 # define isIDFIRST_A(c) generic_isCC_A_(c, CC_IDFIRST_)
1634 # define isALPHA_L1(c) generic_isCC_(c, CC_ALPHA_)
1635 # define isALPHANUMERIC_L1(c) generic_isCC_(c, CC_ALPHANUMERIC_)
1636 # define isBLANK_L1(c) generic_isCC_(c, CC_BLANK_)
1638 /* continuation character for legal NAME in \N{NAME} */
1639 # define isCHARNAME_CONT(c) generic_isCC_(c, CC_CHARNAME_CONT_)
1641 # define isCNTRL_L1(c) generic_isCC_(c, CC_CNTRL_)
1642 # define isGRAPH_L1(c) generic_isCC_(c, CC_GRAPH_)
1643 # define isLOWER_L1(c) generic_isCC_(c, CC_LOWER_)
1644 # define isPRINT_L1(c) generic_isCC_(c, CC_PRINT_)
1645 # define isPSXSPC_L1(c) isSPACE_L1(c)
1646 # define isPUNCT_L1(c) generic_isCC_(c, CC_PUNCT_)
1647 # define isSPACE_L1(c) generic_isCC_(c, CC_SPACE_)
1648 # define isUPPER_L1(c) generic_isCC_(c, CC_UPPER_)
1649 # define isWORDCHAR_L1(c) generic_isCC_(c, CC_WORDCHAR_)
1650 # define isIDFIRST_L1(c) generic_isCC_(c, CC_IDFIRST_)
1653 # define isASCII(c) generic_isCC_(c, CC_ASCII_)
1656 /* Participates in a single-character fold with a character above 255 */
1657 # if defined(PERL_IN_REGCOMP_C) || defined(PERL_IN_REGEXEC_C)
1658 # define HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(c) \
1659 (( ! cBOOL(FITS_IN_8_BITS(c))) \
1660 || (PL_charclass[(U8) (c)] & CC_mask_(CC_NONLATIN1_SIMPLE_FOLD_)))
1662 # define IS_NON_FINAL_FOLD(c) generic_isCC_(c, CC_NON_FINAL_FOLD_)
1663 # define IS_IN_SOME_FOLD_L1(c) generic_isCC_(c, CC_IS_IN_SOME_FOLD_)
1666 /* Like the above, but also can be part of a multi-char fold */
1667 # define HAS_NONLATIN1_FOLD_CLOSURE(c) \
1668 ( (! cBOOL(FITS_IN_8_BITS(c))) \
1669 || (PL_charclass[(U8) (c)] & CC_mask_(CC_NONLATIN1_FOLD_)))
1671 # define _isQUOTEMETA(c) generic_isCC_(c, CC_QUOTEMETA_)
1673 /* is c a control character for which we have a mnemonic? */
1674 # if defined(PERL_CORE) || defined(PERL_EXT)
1675 # define isMNEMONIC_CNTRL(c) generic_isCC_(c, CC_MNEMONIC_CNTRL_)
1677 #else /* else we don't have perl.h H_PERL */
1679 /* If we don't have perl.h, we are compiling a utility program. Below we
1680 * hard-code various macro definitions that wouldn't otherwise be available
1681 * to it. Most are coded based on first principles. These are written to
1682 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
1683 # define isDIGIT_A(c) inRANGE(c, '0', '9')
1684 # define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
1685 # define isSPACE_A(c) (isBLANK_A(c) \
1690 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1691 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1692 * performance (if optimization isn't on), and allow the same code to be
1693 * used for both platform types */
1694 # define isLOWER_A(c) inRANGE((c), 'a', 'i') \
1695 || inRANGE((c), 'j', 'r') \
1696 || inRANGE((c), 's', 'z')
1697 # define isUPPER_A(c) inRANGE((c), 'A', 'I') \
1698 || inRANGE((c), 'J', 'R') \
1699 || inRANGE((c), 'S', 'Z')
1700 # define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1701 # define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
1702 # define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
1703 # define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
1704 # define isXDIGIT_A(c) ( isDIGIT_A(c) \
1705 || inRANGE((c), 'a', 'f') \
1706 || inRANGE((c), 'A', 'F')
1707 # define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1708 || (c) == '#' || (c) == '$' || (c) == '%' \
1709 || (c) == '&' || (c) == '\'' || (c) == '(' \
1710 || (c) == ')' || (c) == '*' || (c) == '+' \
1711 || (c) == ',' || (c) == '.' || (c) == '/' \
1712 || (c) == ':' || (c) == ';' || (c) == '<' \
1713 || (c) == '=' || (c) == '>' || (c) == '?' \
1714 || (c) == '@' || (c) == '[' || (c) == '\\' \
1715 || (c) == ']' || (c) == '^' || (c) == '_' \
1716 || (c) == '`' || (c) == '{' || (c) == '|' \
1717 || (c) == '}' || (c) == '~')
1718 # define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1719 # define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
1722 /* The below is accurate for the 3 EBCDIC code pages traditionally
1723 * supported by perl. The only difference between them in the controls
1724 * is the position of \n, and that is represented symbolically below */
1725 # define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1726 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1727 || (c) == '\t' || (c) == '\v' \
1728 || inRANGE((c), 1, 3) /* SOH, STX, ETX */ \
1729 || (c) == 7F /* U+7F DEL */ \
1730 || inRANGE((c), 0x0E, 0x13) /* SO SI DLE \
1732 || (c) == 0x18 /* U+18 CAN */ \
1733 || (c) == 0x19 /* U+19 EOM */ \
1734 || inRANGE((c), 0x1C, 0x1F) /* [FGRU]S */ \
1735 || (c) == 0x26 /* U+17 ETB */ \
1736 || (c) == 0x27 /* U+1B ESC */ \
1737 || (c) == 0x2D /* U+05 ENQ */ \
1738 || (c) == 0x2E /* U+06 ACK */ \
1739 || (c) == 0x32 /* U+16 SYN */ \
1740 || (c) == 0x37 /* U+04 EOT */ \
1741 || (c) == 0x3C /* U+14 DC4 */ \
1742 || (c) == 0x3D /* U+15 NAK */ \
1743 || (c) == 0x3F)/* U+1A SUB */
1744 # define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
1745 # else /* isASCII is already defined for ASCII platforms, so can use that to
1747 # define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
1750 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
1751 * during debugging, and it seems best to keep them. We may be called
1752 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1753 * do anything anyway, so make it not a problem */
1754 # if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1755 # define NATIVE_TO_LATIN1(ch) (ch)
1757 # define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1758 # define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1759 # define isBLANK_L1(c) (isBLANK_A(c) \
1760 || (FITS_IN_8_BITS(c) \
1761 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1762 # define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1763 # define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1764 # define isLOWER_L1(c) (isLOWER_A(c) \
1765 || (FITS_IN_8_BITS(c) \
1766 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
1767 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1768 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1769 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1770 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1771 # define isPRINT_L1(c) (isPRINT_A(c) \
1772 || (FITS_IN_8_BITS(c) \
1773 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
1774 # define isPUNCT_L1(c) (isPUNCT_A(c) \
1775 || (FITS_IN_8_BITS(c) \
1776 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
1777 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1778 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1779 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1780 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1781 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1782 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1783 # define isSPACE_L1(c) (isSPACE_A(c) \
1784 || (FITS_IN_8_BITS(c) \
1785 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
1786 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1787 # define isUPPER_L1(c) (isUPPER_A(c) \
1788 || (FITS_IN_8_BITS(c) \
1789 && ( IN_RANGE(NATIVE_TO_LATIN1((U8) c), \
1791 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1792 # define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1793 # define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1794 # define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1799 /* The following are not fully accurate in the above-ASCII range. I (khw)
1800 * don't think it's necessary to be so for the purposes where this gets
1802 # define isQUOTEMETA_(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1803 # define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1805 /* And these aren't accurate at all. They are useful only for above
1806 * Latin1, which utilities and bootstrapping don't deal with */
1807 # define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
1808 # define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1809 # define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1811 /* Many of the macros later in this file are defined in terms of these. By
1812 * implementing them with a function, which converts the class number into
1813 * a call to the desired macro, all of the later ones work. However, that
1814 * function won't be actually defined when building a utility program (no
1815 * perl.h), and so a compiler error will be generated if one is attempted
1816 * to be used. And the above-Latin1 code points require Unicode tables to
1817 * be present, something unlikely to be the case when bootstrapping */
1818 # define generic_isCC_(c, classnum) \
1819 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1820 # define generic_isCC_A_(c, classnum) \
1821 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
1822 #endif /* End of no perl.h H_PERL */
1824 #define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1825 #define isALPHA(c) isALPHA_A(c)
1826 #define isASCII_A(c) isASCII(c)
1827 #define isASCII_L1(c) isASCII(c)
1828 #define isBLANK(c) isBLANK_A(c)
1829 #define isCNTRL(c) isCNTRL_A(c)
1830 #define isDIGIT(c) isDIGIT_A(c)
1831 #define isGRAPH(c) isGRAPH_A(c)
1832 #define isIDFIRST(c) isIDFIRST_A(c)
1833 #define isLOWER(c) isLOWER_A(c)
1834 #define isPRINT(c) isPRINT_A(c)
1835 #define isPSXSPC_A(c) isSPACE_A(c)
1836 #define isPSXSPC(c) isPSXSPC_A(c)
1837 #define isPSXSPC_L1(c) isSPACE_L1(c)
1838 #define isPUNCT(c) isPUNCT_A(c)
1839 #define isSPACE(c) isSPACE_A(c)
1840 #define isUPPER(c) isUPPER_A(c)
1841 #define isWORDCHAR(c) isWORDCHAR_A(c)
1842 #define isXDIGIT(c) isXDIGIT_A(c)
1844 /* ASCII casing. These could also be written as
1845 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1846 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1847 which uses table lookup and mask instead of subtraction. (This would
1848 work because the _MOD does not apply in the ASCII range).
1850 These actually are UTF-8 invariant casing, not just ASCII, as any non-ASCII
1851 UTF-8 invariants are neither upper nor lower. (Only on EBCDIC platforms are
1852 there non-ASCII invariants, and all of them are controls.) */
1853 #define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1854 #define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
1856 /* In the ASCII range, these are equivalent to what they're here defined to be.
1857 * But by creating these definitions, other code doesn't have to be aware of
1858 * this detail. Actually this works for all UTF-8 invariants, not just the
1859 * ASCII range. (EBCDIC platforms can have non-ASCII invariants.) */
1860 #define toFOLD(c) toLOWER(c)
1861 #define toTITLE(c) toUPPER(c)
1863 #define toLOWER_A(c) toLOWER(c)
1864 #define toUPPER_A(c) toUPPER(c)
1865 #define toFOLD_A(c) toFOLD(c)
1866 #define toTITLE_A(c) toTITLE(c)
1868 /* Use table lookup for speed; returns the input itself if is out-of-range */
1869 #define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
1871 : PL_latin1_lc[ (U8) (c) ])
1872 #define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1874 /* Modified uc. Is correct uc except for three non-ascii chars which are
1875 * all mapped to one of them, and these need special handling; returns the
1876 * input itself if is out-of-range */
1877 #define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
1879 : PL_mod_latin1_uc[ (U8) (c) ])
1880 #define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
1882 /* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1884 /* For internal core Perl use only: the base macro for defining macros like
1885 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
1886 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1887 * isFOO_L1(); 'classnum' is something like CC_UPPER_, which gives the class
1888 * number for doing this. For non-UTF-8 locales, the code to actually do the
1889 * test this is passed in 'non_utf8'. If 'c' is above 255, 0 is returned. For
1890 * accessing the full range of possible code points under locale rules, use the
1891 * macros based on generic_LC_uvchr_ instead of this. */
1892 #define generic_LC_base_(c, classnum, non_utf8_func) \
1893 (! FITS_IN_8_BITS(c) \
1895 : IN_UTF8_CTYPE_LOCALE \
1896 ? cBOOL(PL_charclass[(U8) (c)] & CC_mask_(classnum)) \
1897 : cBOOL(non_utf8_func(c)))
1899 /* A helper macro for defining macros like isALPHA_LC. On systems without
1900 * proper locales, these reduce to, e.g., isALPHA_A */
1902 # define generic_LC_(c, classnum, non_utf8_func) \
1903 generic_LC_base_(c, classnum, non_utf8_func)
1905 # define generic_LC_(c, classnum, non_utf8_func) \
1906 generic_isCC_A_(c, classnum)
1909 /* Below are the definitions for the locale-sensitive character classification
1910 * macros whose input domain is a byte, and the locale isn't UTF-8. These are
1911 * as close as possible to the bare versions on the platform and still yield
1912 * POSIX Standard-compliant results.
1914 * There is currently only one place these definitions should be used, in
1915 * certain function calls like Perl_iswordchar_() in inline.h.
1917 * Most likely you want to use the macros a ways below with names like
1918 * isALPHA_LC(). Rarely, you may want isU8_ALPHA_LC(), somewhat below.
1920 * The first two aren't in C89, so the fallback is to use the non-locale
1921 * sensitive versions; these are the same for all platforms */
1922 #if defined(HAS_ISASCII)
1923 # define is_base_ASCII(c) isascii((U8) (c))
1925 # define is_base_ASCII(c) isASCII(c)
1928 #if defined(HAS_ISBLANK)
1929 # define is_base_BLANK(c) isblank((U8) (c))
1931 # define is_base_BLANK(c) isBLANK(c)
1934 /* The next few are the same in all platforms. */
1935 #define is_base_CNTRL(c) iscntrl((U8) (c))
1936 #define is_base_IDFIRST(c) (UNLIKELY((c) == '_') || is_base_ALPHA(c))
1937 #define is_base_SPACE(c) isspace((U8) (c))
1938 #define is_base_WORDCHAR(c) (UNLIKELY((c) == '_') || is_base_ALPHANUMERIC(c))
1940 /* The base-level case changing macros are also the same in all platforms */
1941 #define to_base_LOWER(c) tolower((U8) (c))
1942 #define to_base_UPPER(c) toupper((U8) (c))
1943 #define to_base_FOLD(c) to_base_LOWER(c)
1947 /* The Windows functions don't bother to follow the POSIX standard, which for
1948 * example says that something can't both be a printable and a control. But
1949 * Windows treats \t as both a control and a printable, and does such things as
1950 * making superscripts into both digits and punctuation. These #defines tame
1951 * these flaws by assuming that the definitions of controls (and the other few
1952 * ones defined above) are correct, and then making sure that other definitions
1953 * don't have weirdnesses, by adding a check that \w and its subsets aren't
1954 * ispunct(), and things that are \W, like ispunct(), arent't controls. Not
1955 * all possible weirdnesses are checked for, just ones that were detected on
1956 * actual Microsoft code pages */
1957 # define is_base_ALPHA(c) \
1958 (isalpha((U8) (c)) && ! is_base_PUNCT(c))
1959 # define is_base_ALPHANUMERIC(c) \
1960 (isalnum((U8) (c)) && ! is_base_PUNCT(c))
1961 # define is_base_CASED(c) \
1962 ((isupper((U8) (c)) || islower((U8) (c))) && ! is_base_PUNCT(c))
1963 # define is_base_DIGIT(c) \
1964 (isdigit((U8) (c)) && ! is_base_PUNCT(c))
1965 # define is_base_GRAPH(c) \
1966 (isgraph((U8) (c)) && ! is_base_CNTRL(c))
1967 # define is_base_LOWER(c) \
1968 (islower((U8) (c)) && ! is_base_PUNCT(c))
1969 # define is_base_PRINT(c) \
1970 (isprint((U8) (c)) && ! is_base_CNTRL(c))
1971 # define is_base_PUNCT(c) \
1972 (ispunct((U8) (c)) && ! is_base_CNTRL(c))
1973 # define is_base_UPPER(c) \
1974 (isupper((U8) (c)) && ! is_base_PUNCT(c))
1975 # define is_base_XDIGIT(c) \
1976 (isxdigit((U8) (c)) && ! is_base_PUNCT(c))
1979 /* For all other platforms, as far as we know, isdigit(), etc. work sanely
1981 # define is_base_ALPHA(c) isalpha((U8) (c))
1982 # define is_base_ALPHANUMERIC(c) isalnum((U8) (c))
1983 # define is_base_CASED(c) (islower((U8) (c)) || isupper((U8) (c)))
1984 # define is_base_DIGIT(c) isdigit((U8) (c))
1986 /* ... But it seems that IBM products treat NBSP as both a space and a
1987 * graphic; these are the two platforms that we have active test beds for.
1989 # if defined(OS390) || defined(_AIX)
1990 # define is_base_GRAPH(c) (isgraph((U8) (c)) && ! isspace((U8) (c)))
1992 # define is_base_GRAPH(c) isgraph((U8) (c))
1994 # define is_base_LOWER(c) islower((U8) (c))
1995 # define is_base_PRINT(c) isprint((U8) (c))
1996 # define is_base_PUNCT(c) ispunct((U8) (c))
1997 # define is_base_UPPER(c) isupper((U8) (c))
1998 # define is_base_XDIGIT(c) isxdigit((U8) (c))
2001 /* Below is the next level up, which currently expands to nothing more
2002 * than the previous layer. These are the macros to use if you really need
2003 * something whose input domain is a byte, and the locale isn't UTF-8; that is,
2004 * where you normally would have to use things like bare isalnum().
2006 * But most likely you should instead use the layer defined further below which
2007 * has names like isALPHA_LC. They deal with larger-than-byte inputs, and
2010 * (Note, proper general operation of the bare libc functons requires you to
2011 * cast to U8. These do that for you automatically.) */
2013 # define WRAP_U8_LC_(c, classnum, base) base(c)
2015 #define isU8_ALPHANUMERIC_LC(c) \
2016 WRAP_U8_LC_((c), CC_ALPHANUMERIC_, is_base_ALPHANUMERIC)
2017 #define isU8_ALPHA_LC(c) WRAP_U8_LC_((c), CC_ALPHA_, is_base_ALPHA)
2018 #define isU8_ASCII_LC(c) WRAP_U8_LC_((c), CC_ASCII_, is_base_ASCII)
2019 #define isU8_BLANK_LC(c) WRAP_U8_LC_((c), CC_BLANK_, is_base_BLANK)
2020 #define isU8_CASED_LC(c) WRAP_U8_LC_((c), CC_CASED_, is_base_CASED)
2021 #define isU8_CNTRL_LC(c) WRAP_U8_LC_((c), CC_CNTRL_, is_base_CNTRL)
2022 #define isU8_DIGIT_LC(c) WRAP_U8_LC_((c), CC_DIGIT_, is_base_DIGIT)
2023 #define isU8_GRAPH_LC(c) WRAP_U8_LC_((c), CC_GRAPH_, is_base_GRAPH)
2024 #define isU8_IDFIRST_LC(c) WRAP_U8_LC_((c), CC_IDFIRST_, is_base_IDFIRST)
2025 #define isU8_LOWER_LC(c) WRAP_U8_LC_((c), CC_LOWER_, is_base_LOWER)
2026 #define isU8_PRINT_LC(c) WRAP_U8_LC_((c), CC_PRINT_, is_base_PRINT)
2027 #define isU8_PUNCT_LC(c) WRAP_U8_LC_((c), CC_PUNCT_, is_base_PUNCT)
2028 #define isU8_SPACE_LC(c) WRAP_U8_LC_((c), CC_SPACE_, is_base_SPACE)
2029 #define isU8_UPPER_LC(c) WRAP_U8_LC_((c), CC_UPPER_, is_base_UPPER)
2030 #define isU8_WORDCHAR_LC(c) WRAP_U8_LC_((c), CC_WORDCHAR_, is_base_WORDCHAR)
2031 #define isU8_XDIGIT_LC(c) WRAP_U8_LC_((c), CC_XDIGIT_, is_base_XDIGIT)
2033 #define toU8_LOWER_LC(c) WRAP_U8_LC_((c), CC_TOLOWER_, to_base_LOWER)
2034 #define toU8_UPPER_LC(c) WRAP_U8_LC_((c), CC_TOUPPER_, to_base_UPPER)
2035 #define toU8_FOLD_LC(c) toU8_LOWER_LC(c)
2037 /* The definitions below use the ones above to create versions in which the
2038 * input domain isn't restricted to bytes (though always returning false if the
2039 * input doesn't fit in a byte), and to behave properly should the locale be
2040 * UTF-8. These are the documented ones, suitable for general use (though
2041 * toUPPER_LC and toFOLD_LC aren't documented because they need special
2042 * handling to deal with SHARP S expanding to two characters). */
2044 #define isASCII_LC(c) (FITS_IN_8_BITS(c) && isU8_ASCII_LC(c))
2045 #define isALPHA_LC(c) generic_LC_(c, CC_ALPHA_, isU8_ALPHA_LC)
2046 #define isALPHANUMERIC_LC(c) \
2047 generic_LC_(c, CC_ALPHANUMERIC_, isU8_ALPHANUMERIC_LC)
2048 #define isBLANK_LC(c) generic_LC_(c, CC_BLANK_, isU8_BLANK_LC)
2049 #define isCASED_LC(c) generic_LC_(c, CC_CASED_, isU8_CASED_LC)
2050 #define isCNTRL_LC(c) generic_LC_(c, CC_CNTRL_, isU8_CNTRL_LC)
2051 #define isDIGIT_LC(c) generic_LC_(c, CC_DIGIT_, isU8_DIGIT_LC)
2052 #define isGRAPH_LC(c) generic_LC_(c, CC_GRAPH_, isU8_GRAPH_LC)
2053 #define isIDFIRST_LC(c) generic_LC_(c, CC_IDFIRST_, isU8_IDFIRST_LC)
2054 #define isLOWER_LC(c) generic_LC_(c, CC_LOWER_, isU8_LOWER_LC)
2055 #define isPRINT_LC(c) generic_LC_(c, CC_PRINT_, isU8_PRINT_LC)
2056 #define isPUNCT_LC(c) generic_LC_(c, CC_PUNCT_, isU8_PUNCT_LC)
2057 #define isSPACE_LC(c) generic_LC_(c, CC_SPACE_, isU8_SPACE_LC)
2058 #define isUPPER_LC(c) generic_LC_(c, CC_UPPER_, isU8_UPPER_LC)
2059 #define isWORDCHAR_LC(c) generic_LC_(c, CC_WORDCHAR_, isU8_WORDCHAR_LC)
2060 #define isXDIGIT_LC(c) generic_LC_(c, CC_XDIGIT_, isU8_XDIGIT_LC)
2063 # define toLOWER_LC(c) toLOWER_A(c)
2064 # define toUPPER_LC(c) toUPPER_A(c)
2065 # define toFOLD_LC(c) toFOLD_A(c)
2068 /* In the next three macros, the reason for using the PL_latin arrays is in
2069 * case the system function is defective; it ensures uniform results that
2070 * conform to the Unicode standard. */
2072 /* This does not handle the anomalies in UTF-8 Turkic locales. */
2073 # define toLOWER_LC(c) ((! FITS_IN_8_BITS(c)) \
2075 : ((IN_UTF8_CTYPE_LOCALE) \
2076 ? PL_latin1_lc[ (U8) (c) ] \
2077 : ((U8) toU8_LOWER_LC(c))))
2079 /* In this macro, note that the result can be larger than a byte in a UTF-8
2080 * locale. It returns a single value, so can't adequately return the upper
2081 * case of LATIN SMALL LETTER SHARP S in a UTF-8 locale (which should be a
2082 * string of two values "SS"); instead it asserts against that under
2083 * DEBUGGING, and otherwise returns its input. It does not handle the
2084 * anomalies in UTF-8 Turkic locales. */
2085 # define toUPPER_LC(c) \
2086 ((! FITS_IN_8_BITS(c)) \
2088 : ((! IN_UTF8_CTYPE_LOCALE) \
2089 ? ((U8) toU8_UPPER_LC(c)) \
2090 : (UNLIKELY(((U8)(c)) == MICRO_SIGN) \
2091 ? GREEK_CAPITAL_LETTER_MU \
2092 : ((UNLIKELY(((U8) (c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
2093 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
2094 : (UNLIKELY(((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
2095 ? (__ASSERT_(0) (c)) /* Fail on Sharp S in DEBUGGING */ \
2096 : PL_mod_latin1_uc[ (U8) (c) ]))))))
2098 /* In this macro, note that the result can be larger than a byte in a UTF-8
2099 * locale. It returns a single value, so can't adequately return the fold case
2100 * of LATIN SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of
2101 * two values "ss"); instead it asserts against that under DEBUGGING, and
2102 * otherwise returns its input. It does not handle the anomalies in UTF-8
2104 # define toFOLD_LC(c) \
2105 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
2106 ? GREEK_SMALL_LETTER_MU \
2107 : (__ASSERT_( ! IN_UTF8_CTYPE_LOCALE \
2108 || LIKELY((c) != LATIN_SMALL_LETTER_SHARP_S)) \
2112 #define isIDCONT(c) isWORDCHAR(c)
2113 #define isIDCONT_A(c) isWORDCHAR_A(c)
2114 #define isIDCONT_L1(c) isWORDCHAR_L1(c)
2115 #define isIDCONT_LC(c) isWORDCHAR_LC(c)
2116 #define isPSXSPC_LC(c) isSPACE_LC(c)
2118 /* For internal core Perl use only: the base macros for defining macros like
2119 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
2120 * number defined earlier in this file. generic_uvchr_() is used for POSIX
2121 * classes where there is a macro or function 'above_latin1' that takes the
2122 * single argument 'c' and returns the desired value. These exist for those
2123 * classes which have simple definitions, avoiding the overhead of an inversion
2124 * list binary search. generic_invlist_uvchr_() can be used
2125 * for classes where that overhead is faster than a direct lookup.
2126 * generic_uvchr_() won't compile if 'c' isn't unsigned, as it won't match the
2127 * 'above_latin1' prototype. generic_isCC_() macro does bounds checking, so
2128 * have duplicate checks here, so could create versions of the macros that
2129 * don't, but experiments show that gcc optimizes them out anyway. */
2131 /* Note that all ignore 'use bytes' */
2132 #define generic_uvchr_(classnum, above_latin1, c) ((c) < 256 \
2133 ? generic_isCC_(c, classnum) \
2135 #define generic_invlist_uvchr_(classnum, c) ((c) < 256 \
2136 ? generic_isCC_(c, classnum) \
2137 : _is_uni_FOO(classnum, c))
2138 #define isALPHA_uvchr(c) generic_invlist_uvchr_(CC_ALPHA_, c)
2139 #define isALPHANUMERIC_uvchr(c) generic_invlist_uvchr_(CC_ALPHANUMERIC_, c)
2140 #define isASCII_uvchr(c) isASCII(c)
2141 #define isBLANK_uvchr(c) generic_uvchr_(CC_BLANK_, is_HORIZWS_cp_high, c)
2142 #define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
2143 #define isDIGIT_uvchr(c) generic_invlist_uvchr_(CC_DIGIT_, c)
2144 #define isGRAPH_uvchr(c) generic_invlist_uvchr_(CC_GRAPH_, c)
2145 #define isIDCONT_uvchr(c) \
2146 generic_uvchr_(CC_WORDCHAR_, _is_uni_perl_idcont, c)
2147 #define isIDFIRST_uvchr(c) \
2148 generic_uvchr_(CC_IDFIRST_, _is_uni_perl_idstart, c)
2149 #define isLOWER_uvchr(c) generic_invlist_uvchr_(CC_LOWER_, c)
2150 #define isPRINT_uvchr(c) generic_invlist_uvchr_(CC_PRINT_, c)
2152 #define isPUNCT_uvchr(c) generic_invlist_uvchr_(CC_PUNCT_, c)
2153 #define isSPACE_uvchr(c) generic_uvchr_(CC_SPACE_, is_XPERLSPACE_cp_high, c)
2154 #define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
2156 #define isUPPER_uvchr(c) generic_invlist_uvchr_(CC_UPPER_, c)
2157 #define isVERTWS_uvchr(c) generic_uvchr_(CC_VERTSPACE_, is_VERTWS_cp_high, c)
2158 #define isWORDCHAR_uvchr(c) generic_invlist_uvchr_(CC_WORDCHAR_, c)
2159 #define isXDIGIT_uvchr(c) generic_uvchr_(CC_XDIGIT_, is_XDIGIT_cp_high, c)
2161 #define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
2162 #define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
2163 #define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
2164 #define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
2166 /* For backwards compatibility, even though '_uni' should mean official Unicode
2167 * code points, in Perl it means native for those below 256 */
2168 #define isALPHA_uni(c) isALPHA_uvchr(c)
2169 #define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
2170 #define isASCII_uni(c) isASCII_uvchr(c)
2171 #define isBLANK_uni(c) isBLANK_uvchr(c)
2172 #define isCNTRL_uni(c) isCNTRL_uvchr(c)
2173 #define isDIGIT_uni(c) isDIGIT_uvchr(c)
2174 #define isGRAPH_uni(c) isGRAPH_uvchr(c)
2175 #define isIDCONT_uni(c) isIDCONT_uvchr(c)
2176 #define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
2177 #define isLOWER_uni(c) isLOWER_uvchr(c)
2178 #define isPRINT_uni(c) isPRINT_uvchr(c)
2179 #define isPUNCT_uni(c) isPUNCT_uvchr(c)
2180 #define isSPACE_uni(c) isSPACE_uvchr(c)
2181 #define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
2182 #define isUPPER_uni(c) isUPPER_uvchr(c)
2183 #define isVERTWS_uni(c) isVERTWS_uvchr(c)
2184 #define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
2185 #define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
2186 #define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
2187 #define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
2188 #define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
2189 #define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
2191 /* For internal core Perl use only: the base macros for defining macros like
2192 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
2193 * point, not just 0-255. Like generic_uvchr_, there are two versions, one for
2194 * simple class definitions; the other for more complex. These are like
2195 * generic_uvchr_, so see it for more info. */
2196 #define generic_LC_uvchr_(latin1, above_latin1, c) \
2197 (c < 256 ? latin1(c) : above_latin1(c))
2198 #define generic_LC_invlist_uvchr_(latin1, classnum, c) \
2199 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
2201 #define isALPHA_LC_uvchr(c) generic_LC_invlist_uvchr_(isALPHA_LC, CC_ALPHA_, c)
2202 #define isALPHANUMERIC_LC_uvchr(c) generic_LC_invlist_uvchr_(isALPHANUMERIC_LC, \
2203 CC_ALPHANUMERIC_, c)
2204 #define isASCII_LC_uvchr(c) isASCII_LC(c)
2205 #define isBLANK_LC_uvchr(c) generic_LC_uvchr_(isBLANK_LC, \
2206 is_HORIZWS_cp_high, c)
2207 #define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
2208 #define isDIGIT_LC_uvchr(c) generic_LC_invlist_uvchr_(isDIGIT_LC, CC_DIGIT_, c)
2209 #define isGRAPH_LC_uvchr(c) generic_LC_invlist_uvchr_(isGRAPH_LC, CC_GRAPH_, c)
2210 #define isIDCONT_LC_uvchr(c) generic_LC_uvchr_(isIDCONT_LC, \
2211 _is_uni_perl_idcont, c)
2212 #define isIDFIRST_LC_uvchr(c) generic_LC_uvchr_(isIDFIRST_LC, \
2213 _is_uni_perl_idstart, c)
2214 #define isLOWER_LC_uvchr(c) generic_LC_invlist_uvchr_(isLOWER_LC, CC_LOWER_, c)
2215 #define isPRINT_LC_uvchr(c) generic_LC_invlist_uvchr_(isPRINT_LC, CC_PRINT_, c)
2216 #define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
2217 #define isPUNCT_LC_uvchr(c) generic_LC_invlist_uvchr_(isPUNCT_LC, CC_PUNCT_, c)
2218 #define isSPACE_LC_uvchr(c) generic_LC_uvchr_(isSPACE_LC, \
2219 is_XPERLSPACE_cp_high, c)
2220 #define isUPPER_LC_uvchr(c) generic_LC_invlist_uvchr_(isUPPER_LC, CC_UPPER_, c)
2221 #define isWORDCHAR_LC_uvchr(c) generic_LC_invlist_uvchr_(isWORDCHAR_LC, \
2223 #define isXDIGIT_LC_uvchr(c) generic_LC_uvchr_(isXDIGIT_LC, \
2224 is_XDIGIT_cp_high, c)
2226 #define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
2228 /* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
2229 * they don't otherwise go out of their way to look for malformed UTF-8. If
2230 * they can return accurate results without knowing if the input is otherwise
2231 * malformed, they do so. For example isASCII is accurate in spite of any
2232 * non-length malformations because it looks only at a single byte. Likewise
2233 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
2234 * variant ones return FALSE. But, if the input has to be well-formed in order
2235 * for the results to be accurate, the macros will test and if malformed will
2236 * call a routine to die
2238 * Except for toke.c, the macros do assume that e > p, asserting that on
2239 * DEBUGGING builds. Much code that calls these depends on this being true,
2240 * for other reasons. toke.c is treated specially as using the regular
2241 * assertion breaks it in many ways. All strings that these operate on there
2242 * are supposed to have an extra NUL character at the end, so that *e = \0. A
2243 * bunch of code in toke.c assumes that this is true, so the assertion allows
2245 #ifdef PERL_IN_TOKE_C
2246 # define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
2248 # define _utf8_safe_assert(p,e) ((e) > (p))
2251 #define generic_utf8_safe_(classnum, p, e, above_latin1) \
2252 ((! _utf8_safe_assert(p, e)) \
2253 ? (_force_out_malformed_utf8_message((U8 *) (p), (U8 *) (e), 0, 1), 0)\
2254 : (UTF8_IS_INVARIANT(*(p))) \
2255 ? generic_isCC_(*(p), classnum) \
2256 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2257 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2258 ? generic_isCC_(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
2260 : (_force_out_malformed_utf8_message( \
2261 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2263 /* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
2264 * 'above_latin1' can be a macro */
2265 #define generic_func_utf8_safe_(classnum, above_latin1, p, e) \
2266 generic_utf8_safe_(classnum, p, e, above_latin1(p, e))
2267 #define generic_non_invlist_utf8_safe_(classnum, above_latin1, p, e) \
2268 generic_utf8_safe_(classnum, p, e, \
2269 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2270 ? (_force_out_malformed_utf8_message( \
2271 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2273 /* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
2274 * 'above_latin1' parameter */
2275 #define generic_invlist_utf8_safe_(classnum, p, e) \
2276 generic_utf8_safe_(classnum, p, e, _is_utf8_FOO(classnum, p, e))
2278 /* Like the above, but should be used only when it is known that there are no
2279 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
2280 * class is TRUE for. Hence it can skip the tests for this range.
2281 * 'above_latin1' should include its arguments */
2282 #define generic_utf8_safe_no_upper_latin1_(classnum, p, e, above_latin1) \
2283 (__ASSERT_(_utf8_safe_assert(p, e)) \
2285 ? generic_isCC_(*(p), classnum) \
2286 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
2287 ? 0 /* Note that doesn't check validity for latin1 */ \
2291 #define isALPHA_utf8(p, e) isALPHA_utf8_safe(p, e)
2292 #define isALPHANUMERIC_utf8(p, e) isALPHANUMERIC_utf8_safe(p, e)
2293 #define isASCII_utf8(p, e) isASCII_utf8_safe(p, e)
2294 #define isBLANK_utf8(p, e) isBLANK_utf8_safe(p, e)
2295 #define isCNTRL_utf8(p, e) isCNTRL_utf8_safe(p, e)
2296 #define isDIGIT_utf8(p, e) isDIGIT_utf8_safe(p, e)
2297 #define isGRAPH_utf8(p, e) isGRAPH_utf8_safe(p, e)
2298 #define isIDCONT_utf8(p, e) isIDCONT_utf8_safe(p, e)
2299 #define isIDFIRST_utf8(p, e) isIDFIRST_utf8_safe(p, e)
2300 #define isLOWER_utf8(p, e) isLOWER_utf8_safe(p, e)
2301 #define isPRINT_utf8(p, e) isPRINT_utf8_safe(p, e)
2302 #define isPSXSPC_utf8(p, e) isPSXSPC_utf8_safe(p, e)
2303 #define isPUNCT_utf8(p, e) isPUNCT_utf8_safe(p, e)
2304 #define isSPACE_utf8(p, e) isSPACE_utf8_safe(p, e)
2305 #define isUPPER_utf8(p, e) isUPPER_utf8_safe(p, e)
2306 #define isVERTWS_utf8(p, e) isVERTWS_utf8_safe(p, e)
2307 #define isWORDCHAR_utf8(p, e) isWORDCHAR_utf8_safe(p, e)
2308 #define isXDIGIT_utf8(p, e) isXDIGIT_utf8_safe(p, e)
2310 #define isALPHA_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_ALPHA_, p, e)
2311 #define isALPHANUMERIC_utf8_safe(p, e) \
2312 generic_invlist_utf8_safe_(CC_ALPHANUMERIC_, p, e)
2313 #define isASCII_utf8_safe(p, e) \
2314 /* Because ASCII is invariant under utf8, the non-utf8 macro \
2316 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
2317 #define isBLANK_utf8_safe(p, e) \
2318 generic_non_invlist_utf8_safe_(CC_BLANK_, is_HORIZWS_high, p, e)
2321 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
2322 * more efficient macro instead of the more general one */
2323 # define isCNTRL_utf8_safe(p, e) \
2324 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
2326 # define isCNTRL_utf8_safe(p, e) generic_utf8_safe_(CC_CNTRL_, p, e, 0)
2329 #define isDIGIT_utf8_safe(p, e) \
2330 generic_utf8_safe_no_upper_latin1_(CC_DIGIT_, p, e, \
2331 _is_utf8_FOO(CC_DIGIT_, p, e))
2332 #define isGRAPH_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_GRAPH_, p, e)
2333 #define isIDCONT_utf8_safe(p, e) generic_func_utf8_safe_(CC_WORDCHAR_, \
2334 _is_utf8_perl_idcont, p, e)
2336 /* To prevent S_scan_word in toke.c from hanging, we have to make sure that
2337 * IDFIRST is an alnum. See
2338 * https://github.com/Perl/perl5/issues/10275 for more detail than you
2339 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
2340 * This used to be not the XID version, but we decided to go with the more
2341 * modern Unicode definition */
2342 #define isIDFIRST_utf8_safe(p, e) \
2343 generic_func_utf8_safe_(CC_IDFIRST_, \
2344 _is_utf8_perl_idstart, (U8 *) (p), (U8 *) (e))
2346 #define isLOWER_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_LOWER_, p, e)
2347 #define isPRINT_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_PRINT_, p, e)
2348 #define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
2349 #define isPUNCT_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_PUNCT_, p, e)
2350 #define isSPACE_utf8_safe(p, e) \
2351 generic_non_invlist_utf8_safe_(CC_SPACE_, is_XPERLSPACE_high, p, e)
2352 #define isUPPER_utf8_safe(p, e) generic_invlist_utf8_safe_(CC_UPPER_, p, e)
2353 #define isVERTWS_utf8_safe(p, e) \
2354 generic_non_invlist_utf8_safe_(CC_VERTSPACE_, is_VERTWS_high, p, e)
2355 #define isWORDCHAR_utf8_safe(p, e) \
2356 generic_invlist_utf8_safe_(CC_WORDCHAR_, p, e)
2357 #define isXDIGIT_utf8_safe(p, e) \
2358 generic_utf8_safe_no_upper_latin1_(CC_XDIGIT_, p, e, \
2359 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2360 ? (_force_out_malformed_utf8_message( \
2361 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2362 : is_XDIGIT_high(p)))
2364 #define toFOLD_utf8(p,e,s,l) toFOLD_utf8_safe(p,e,s,l)
2365 #define toLOWER_utf8(p,e,s,l) toLOWER_utf8_safe(p,e,s,l)
2366 #define toTITLE_utf8(p,e,s,l) toTITLE_utf8_safe(p,e,s,l)
2367 #define toUPPER_utf8(p,e,s,l) toUPPER_utf8_safe(p,e,s,l)
2369 /* For internal core use only, subject to change */
2370 #define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f)
2371 #define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f)
2372 #define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f)
2373 #define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f)
2375 #define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
2376 #define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
2377 #define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
2378 #define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
2380 #define isALPHA_LC_utf8(p, e) isALPHA_LC_utf8_safe(p, e)
2381 #define isALPHANUMERIC_LC_utf8(p, e) isALPHANUMERIC_LC_utf8_safe(p, e)
2382 #define isASCII_LC_utf8(p, e) isASCII_LC_utf8_safe(p, e)
2383 #define isBLANK_LC_utf8(p, e) isBLANK_LC_utf8_safe(p, e)
2384 #define isCNTRL_LC_utf8(p, e) isCNTRL_LC_utf8_safe(p, e)
2385 #define isDIGIT_LC_utf8(p, e) isDIGIT_LC_utf8_safe(p, e)
2386 #define isGRAPH_LC_utf8(p, e) isGRAPH_LC_utf8_safe(p, e)
2387 #define isIDCONT_LC_utf8(p, e) isIDCONT_LC_utf8_safe(p, e)
2388 #define isIDFIRST_LC_utf8(p, e) isIDFIRST_LC_utf8_safe(p, e)
2389 #define isLOWER_LC_utf8(p, e) isLOWER_LC_utf8_safe(p, e)
2390 #define isPRINT_LC_utf8(p, e) isPRINT_LC_utf8_safe(p, e)
2391 #define isPSXSPC_LC_utf8(p, e) isPSXSPC_LC_utf8_safe(p, e)
2392 #define isPUNCT_LC_utf8(p, e) isPUNCT_LC_utf8_safe(p, e)
2393 #define isSPACE_LC_utf8(p, e) isSPACE_LC_utf8_safe(p, e)
2394 #define isUPPER_LC_utf8(p, e) isUPPER_LC_utf8_safe(p, e)
2395 #define isWORDCHAR_LC_utf8(p, e) isWORDCHAR_LC_utf8_safe(p, e)
2396 #define isXDIGIT_LC_utf8(p, e) isXDIGIT_LC_utf8_safe(p, e)
2398 /* For internal core Perl use only: the base macros for defining macros like
2399 * isALPHA_LC_utf8_safe. These are like generic_utf8_, but if the first code
2400 * point in 'p' is within the 0-255 range, it uses locale rules from the
2401 * passed-in 'macro' parameter */
2402 #define generic_LC_utf8_safe_(macro, p, e, above_latin1) \
2403 (__ASSERT_(_utf8_safe_assert(p, e)) \
2404 (UTF8_IS_INVARIANT(*(p))) \
2406 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2407 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2408 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2409 : (_force_out_malformed_utf8_message( \
2410 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2413 #define generic_LC_invlist_utf8_safe_(macro, classnum, p, e) \
2414 generic_LC_utf8_safe_(macro, p, e, \
2415 _is_utf8_FOO(classnum, p, e))
2417 #define generic_LC_func_utf8_safe_(macro, above_latin1, p, e) \
2418 generic_LC_utf8_safe_(macro, p, e, above_latin1(p, e))
2420 #define generic_LC_non_invlist_utf8_safe_(classnum, above_latin1, p, e) \
2421 generic_LC_utf8_safe_(classnum, p, e, \
2422 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2423 ? (_force_out_malformed_utf8_message( \
2424 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2427 #define isALPHANUMERIC_LC_utf8_safe(p, e) \
2428 generic_LC_invlist_utf8_safe_(isALPHANUMERIC_LC, \
2429 CC_ALPHANUMERIC_, p, e)
2430 #define isALPHA_LC_utf8_safe(p, e) \
2431 generic_LC_invlist_utf8_safe_(isALPHA_LC, CC_ALPHA_, p, e)
2432 #define isASCII_LC_utf8_safe(p, e) \
2433 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2434 #define isBLANK_LC_utf8_safe(p, e) \
2435 generic_LC_non_invlist_utf8_safe_(isBLANK_LC, is_HORIZWS_high, p, e)
2436 #define isCNTRL_LC_utf8_safe(p, e) \
2437 generic_LC_utf8_safe_(isCNTRL_LC, p, e, 0)
2438 #define isDIGIT_LC_utf8_safe(p, e) \
2439 generic_LC_invlist_utf8_safe_(isDIGIT_LC, CC_DIGIT_, p, e)
2440 #define isGRAPH_LC_utf8_safe(p, e) \
2441 generic_LC_invlist_utf8_safe_(isGRAPH_LC, CC_GRAPH_, p, e)
2442 #define isIDCONT_LC_utf8_safe(p, e) \
2443 generic_LC_func_utf8_safe_(isIDCONT_LC, \
2444 _is_utf8_perl_idcont, p, e)
2445 #define isIDFIRST_LC_utf8_safe(p, e) \
2446 generic_LC_func_utf8_safe_(isIDFIRST_LC, \
2447 _is_utf8_perl_idstart, p, e)
2448 #define isLOWER_LC_utf8_safe(p, e) \
2449 generic_LC_invlist_utf8_safe_(isLOWER_LC, CC_LOWER_, p, e)
2450 #define isPRINT_LC_utf8_safe(p, e) \
2451 generic_LC_invlist_utf8_safe_(isPRINT_LC, CC_PRINT_, p, e)
2452 #define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2453 #define isPUNCT_LC_utf8_safe(p, e) \
2454 generic_LC_invlist_utf8_safe_(isPUNCT_LC, CC_PUNCT_, p, e)
2455 #define isSPACE_LC_utf8_safe(p, e) \
2456 generic_LC_non_invlist_utf8_safe_(isSPACE_LC, is_XPERLSPACE_high, p, e)
2457 #define isUPPER_LC_utf8_safe(p, e) \
2458 generic_LC_invlist_utf8_safe_(isUPPER_LC, CC_UPPER_, p, e)
2459 #define isWORDCHAR_LC_utf8_safe(p, e) \
2460 generic_LC_invlist_utf8_safe_(isWORDCHAR_LC, CC_WORDCHAR_, p, e)
2461 #define isXDIGIT_LC_utf8_safe(p, e) \
2462 generic_LC_non_invlist_utf8_safe_(isXDIGIT_LC, is_XDIGIT_high, p, e)
2464 /* Macros for backwards compatibility and for completeness when the ASCII and
2465 * Latin1 values are identical */
2466 #define isALPHAU(c) isALPHA_L1(c)
2467 #define isDIGIT_L1(c) isDIGIT_A(c)
2468 #define isOCTAL(c) isOCTAL_A(c)
2469 #define isOCTAL_L1(c) isOCTAL_A(c)
2470 #define isXDIGIT_L1(c) isXDIGIT_A(c)
2471 #define isALNUM(c) isWORDCHAR(c)
2472 #define isALNUM_A(c) isALNUM(c)
2473 #define isALNUMU(c) isWORDCHAR_L1(c)
2474 #define isALNUM_LC(c) isWORDCHAR_LC(c)
2475 #define isALNUM_uni(c) isWORDCHAR_uni(c)
2476 #define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
2477 #define isALNUM_utf8(p,e) isWORDCHAR_utf8(p,e)
2478 #define isALNUM_utf8_safe(p,e) isWORDCHAR_utf8_safe(p,e)
2479 #define isALNUM_LC_utf8(p,e)isWORDCHAR_LC_utf8(p,e)
2480 #define isALNUM_LC_utf8_safe(p,e)isWORDCHAR_LC_utf8_safe(p,e)
2481 #define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2482 #define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2483 #define isALNUMC(c) isALPHANUMERIC(c)
2484 #define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2485 #define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
2486 #define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
2487 #define isALNUMC_utf8(p,e) isALPHANUMERIC_utf8(p,e)
2488 #define isALNUMC_utf8_safe(p,e) isALPHANUMERIC_utf8_safe(p,e)
2489 #define isALNUMC_LC_utf8_safe(p,e) isALPHANUMERIC_LC_utf8_safe(p,e)
2491 /* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2492 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2493 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2494 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2495 * the outlier from the block that contains the other controls, just like
2496 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2497 * block. If it weren't special cased, it would yield a non-control.
2498 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2501 # define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2503 # define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2505 ? (UNLIKELY((c) == '?') \
2506 ? QUESTION_MARK_CTRL \
2507 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2508 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2510 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2514 =for apidoc Ay||line_t
2515 The typedef to use to declare variables that are to hold line numbers.
2519 Line numbers are unsigned, 32 bits.
2522 #define LINE_Tf U32uf
2523 #define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
2525 /* Helpful alias for version prescan */
2526 #define is_LAX_VERSION(a,b) \
2527 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2529 #define is_STRICT_VERSION(a,b) \
2530 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2532 #define BADVERSION(a,b,c) \
2538 /* Converts a character KNOWN to represent a hexadecimal digit (0-9, A-F, or
2539 * a-f) to its numeric value without using any branches. The input is
2540 * validated only by an assert() in DEBUGGING builds.
2542 * It works by right shifting and isolating the bit that is 0 for the digits,
2543 * and 1 for at least the alphas A-F, a-f. The bit is shifted to the ones
2544 * position, and then to the eights position. Both are added together to form
2545 * 0 if the input is '0'-'9' and to form 9 if alpha. This is added to the
2546 * final four bits of the input to form the correct value. */
2547 #define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) \
2548 ((NATIVE_TO_LATIN1(c) >> 6) & 1) /* 1 if alpha; 0 if not */ \
2549 + ((NATIVE_TO_LATIN1(c) >> 3) & 8) /* 8 if alpha; 0 if not */ \
2550 + ((c) & 0xF)) /* 0-9 if input valid hex digit */
2552 /* The argument is a string pointer, which is advanced. */
2553 #define READ_XDIGIT(s) ((s)++, XDIGIT_VALUE(*((s) - 1)))
2555 /* Converts a character known to represent an octal digit (0-7) to its numeric
2556 * value. The input is validated only by an assert() in DEBUGGING builds. In
2557 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2558 #define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2560 /* Efficiently returns a boolean as to if two native characters are equivalent
2561 * case-insensitively. At least one of the characters must be one of [A-Za-z];
2562 * the ALPHA in the name is to remind you of that. This is asserted() in
2563 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2564 * works (on valid input) for both non- and UTF-8-encoded bytes.
2566 * When one of the inputs is a compile-time constant and gets folded by the
2567 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2568 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2569 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
2570 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2571 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2572 * just a single 0, in the bit position where the upper- and lowercase differ.
2574 #define isALPHA_FOLD_EQ(c1, c2) \
2575 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2576 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2577 #define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2580 =for apidoc_section $memory
2582 =for apidoc Am|void|Newx|void* ptr|int nitems|type
2583 =for apidoc_item |void*|safemalloc|size_t size
2585 The XSUB-writer's interface to the C C<malloc> function.
2587 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2589 In 5.9.3, Newx() and friends replace the older New() API, and drops
2590 the first parameter, I<x>, a debug aid which allowed callers to identify
2591 themselves. This aid has been superseded by a new build option,
2592 PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
2593 there for use in XS modules supporting older perls.
2595 =for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
2596 The XSUB-writer's interface to the C C<malloc> function, with
2597 cast. See also C<L</Newx>>.
2599 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2601 =for apidoc Am|void|Newxz|void* ptr|int nitems|type
2602 =for apidoc_item |void*|safecalloc|size_t nitems|size_t item_size
2604 The XSUB-writer's interface to the C C<malloc> function. The allocated
2605 memory is zeroed with C<memzero>. See also C<L</Newx>>.
2607 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2609 =for apidoc Am|void|Renew|void* ptr|int nitems|type
2610 =for apidoc_item |void*|saferealloc|void *ptr|size_t size
2612 The XSUB-writer's interface to the C C<realloc> function.
2614 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2616 =for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2617 The XSUB-writer's interface to the C C<realloc> function, with
2620 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2622 =for apidoc Am|void|Safefree|void* ptr
2623 The XSUB-writer's interface to the C C<free> function.
2625 This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
2627 =for apidoc_section $string
2628 =for apidoc Am|void |Move |void* src|void* dest|int nitems|type
2629 =for apidoc_item |void *|MoveD|void* src|void* dest|int nitems|type
2630 The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
2631 source, C<dest> is the destination, C<nitems> is the number of items, and
2632 C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
2634 C<MoveD> is like C<Move> but returns C<dest>. Useful
2635 for encouraging compilers to tail-call
2638 =for apidoc Am|void |Copy |void* src|void* dest|int nitems|type
2639 =for apidoc_item |void *|CopyD|void* src|void* dest|int nitems|type
2640 The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
2641 source, C<dest> is the destination, C<nitems> is the number of items, and
2642 C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
2644 C<CopyD> is like C<Copy> but returns C<dest>. Useful
2645 for encouraging compilers to tail-call
2648 =for apidoc Am|void |Zero |void* dest|int nitems|type
2649 =for apidoc_item |void *|ZeroD|void* dest|int nitems|type
2651 The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2652 destination, C<nitems> is the number of items, and C<type> is the type.
2654 C<ZeroD> is like C<Zero> but returns C<dest>. Useful
2655 for encouraging compilers to tail-call
2658 =for apidoc_section $utility
2659 =for apidoc Amu|void|StructCopy|type *src|type *dest|type
2660 This is an architecture-independent macro to copy one structure to another.
2662 =for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2664 Fill up memory with a byte pattern (a byte repeated over and over
2665 again) that hopefully catches attempts to access uninitialized memory.
2667 =for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2669 PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2671 =for apidoc Am|void|PoisonFree|void* dest|int nitems|type
2673 PoisonWith(0xEF) for catching access to freed memory.
2675 =for apidoc Am|void|Poison|void* dest|int nitems|type
2677 PoisonWith(0xEF) for catching access to freed memory.
2681 /* Maintained for backwards-compatibility only. Use newSV() instead. */
2683 #define NEWSV(x,len) newSV(len)
2686 #define MEM_SIZE_MAX ((MEM_SIZE)-1)
2688 #define _PERL_STRLEN_ROUNDUP_UNCHECKED(n) (((n) - 1 + PERL_STRLEN_ROUNDUP_QUANTUM) & ~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM - 1))
2690 #ifdef PERL_MALLOC_WRAP
2692 /* This expression will be constant-folded at compile time. It checks
2693 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2694 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2695 * As well as avoiding the need for a run-time check in some cases, it's
2696 * designed to avoid compiler warnings like:
2697 * comparison is always false due to limited range of data type
2698 * It's mathematically equivalent to
2699 * max(n) * sizeof(t) > MEM_SIZE_MAX
2702 # define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
2703 ( sizeof(MEM_SIZE) < sizeof(n) \
2704 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
2706 /* This is written in a slightly odd way to avoid various spurious
2707 * compiler warnings. We *want* to write the expression as
2708 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2709 * (for some compile-time constant C), but even when the LHS
2710 * constant-folds to false at compile-time, g++ insists on emitting
2711 * warnings about the RHS (e.g. "comparison is always false"), so instead
2714 * (cond ? n : X) > C
2716 * where X is a constant with X > C always false. Choosing a value for X
2717 * is tricky. If 0, some compilers will complain about 0 > C always being
2718 * false; if 1, Coverity complains when n happens to be the constant value
2719 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2720 * for X and hope that nothing else whines.
2723 # define _MEM_WRAP_WILL_WRAP(n,t) \
2724 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2725 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
2727 # define MEM_WRAP_CHECK(n,t) \
2728 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2729 && (croak_memory_wrap(),0))
2731 # define MEM_WRAP_CHECK_1(n,t,a) \
2732 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2733 && (Perl_croak_nocontext("%s",(a)),0))
2735 /* "a" arg must be a string literal */
2736 # define MEM_WRAP_CHECK_s(n,t,a) \
2737 ( (void) (UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2738 && (Perl_croak_nocontext(ASSERT_IS_LITERAL(a)), 0)))
2740 # define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
2742 # define PERL_STRLEN_ROUNDUP(n) ((void)(((n) > MEM_SIZE_MAX - 2 * PERL_STRLEN_ROUNDUP_QUANTUM) ? (croak_memory_wrap(),0) : 0), _PERL_STRLEN_ROUNDUP_UNCHECKED(n))
2745 # define MEM_WRAP_CHECK(n,t)
2746 # define MEM_WRAP_CHECK_1(n,t,a)
2747 # define MEM_WRAP_CHECK_s(n,t,a)
2748 # define MEM_WRAP_CHECK_(n,t)
2750 # define PERL_STRLEN_ROUNDUP(n) _PERL_STRLEN_ROUNDUP_UNCHECKED(n)
2756 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
2757 * go through functions, which are handy for debugging breakpoints, but
2758 * which more importantly get the immediate calling environment (file and
2759 * line number, and C function name if available) passed in. This info can
2760 * then be used for logging the calls, for which one gets a sample
2761 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
2764 * - not all memory allocs get logged, only those
2765 * that go through Newx() and derivatives (while all
2766 * Safefrees do get logged)
2767 * - __FILE__ and __LINE__ do not work everywhere
2768 * - __func__ or __FUNCTION__ even less so
2769 * - I think more goes on after the perlio frees but
2770 * the thing is that STDERR gets closed (as do all
2771 * the file descriptors)
2772 * - no deeper calling stack than the caller of the Newx()
2773 * or the kind, but do I look like a C reflection/introspection
2775 * - the function prototypes for the logging functions
2776 * probably should maybe be somewhere else than handy.h
2777 * - one could consider inlining (macrofying) the logging
2778 * for speed, but I am too lazy
2779 * - one could imagine recording the allocations in a hash,
2780 * (keyed by the allocation address?), and maintain that
2781 * through reallocs and frees, but how to do that without
2782 * any News() happening...?
2783 * - lots of -Ddefines to get useful/controllable output
2784 * - lots of ENV reads
2788 # ifndef PERL_MEM_LOG_NOIMPL
2802 #define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2803 #define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
2804 #define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
2807 #ifndef MEM_LOG_ALLOC
2808 #define MEM_LOG_ALLOC(n,t,a) (a)
2810 #ifndef MEM_LOG_REALLOC
2811 #define MEM_LOG_REALLOC(n,t,v,a) (a)
2813 #ifndef MEM_LOG_FREE
2814 #define MEM_LOG_FREE(a) (a)
2817 #define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2818 #define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2819 #define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
2822 /* pre 5.9.x compatibility */
2823 #define New(x,v,n,t) Newx(v,n,t)
2824 #define Newc(x,v,n,t,c) Newxc(v,n,t,c)
2825 #define Newz(x,v,n,t) Newxz(v,n,t)
2828 #define Renew(v,n,t) \
2829 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
2830 #define Renewc(v,n,t,c) \
2831 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
2834 #define Safefree(d) \
2835 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
2837 #define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
2840 /* assert that a valid ptr has been supplied - use this instead of assert(ptr) *
2841 * as it handles cases like constant string arguments without throwing warnings *
2842 * the cast is required, as is the inequality check, to avoid warnings */
2843 #define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
2846 #define Move(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), (void)memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2847 #define Copy(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), (void)memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
2848 #define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2850 /* Like above, but returns a pointer to 'd' */
2851 #define MoveD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2852 #define CopyD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
2853 #define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
2855 #define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2856 #define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2857 #define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2858 #define Poison(d,n,t) PoisonFree(d,n,t)
2861 # define PERL_POISON_EXPR(x) x
2863 # define PERL_POISON_EXPR(x)
2867 #define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2870 =for apidoc_section $utility
2872 =for apidoc Am|STRLEN|C_ARRAY_LENGTH|void *a
2874 Returns the number of elements in the input C array (so you want your
2875 zero-based indices to be less than but not equal to).
2877 =for apidoc Am|void *|C_ARRAY_END|void *a
2879 Returns a pointer to one element past the final element of the input C array.
2883 C_ARRAY_END is one past the last: half-open/half-closed range, not
2884 last-inclusive range.
2886 #define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
2887 #define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
2889 #if defined(PERL_CORE) || defined(PERL_EXT_RE_BUILD)
2890 /* strlen() of a literal string constant. Restricting this to core, in part
2891 * because it can generate compiler warnings about comparing unlike signs */
2892 # define STRLENs(s) (sizeof("" s "") - 1)
2897 # define Perl_va_copy(s, d) va_copy(d, s)
2898 # elif defined(__va_copy)
2899 # define Perl_va_copy(s, d) __va_copy(d, s)
2901 # define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2905 /* convenience debug macros */
2907 #define pTHX_FORMAT "Perl interpreter: 0x%p"
2908 #define pTHX__FORMAT ", Perl interpreter: 0x%p"
2909 #define pTHX_VALUE_ (void *)my_perl,
2910 #define pTHX_VALUE (void *)my_perl
2911 #define pTHX__VALUE_ ,(void *)my_perl,
2912 #define pTHX__VALUE ,(void *)my_perl
2915 #define pTHX__FORMAT
2918 #define pTHX__VALUE_
2920 #endif /* USE_ITHREADS */
2922 /* Perl_deprecate was not part of the public API, and did not have a deprecate()
2923 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2924 CPAN::Unpack show any users outside the core. */
2926 # define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2927 "Use of " s " is deprecated")
2928 # define deprecate_disappears_in(when,message) \
2929 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2930 message " is deprecated, and will disappear in Perl " when)
2931 # define deprecate_fatal_in(when,message) \
2932 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2933 message " is deprecated, and will become fatal in Perl " when)
2936 /* Internal macros to deal with gids and uids */
2939 # if Uid_t_size > IVSIZE
2940 # define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2941 # define SvUID(sv) SvNV(sv)
2942 # elif Uid_t_sign <= 0
2943 # define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2944 # define SvUID(sv) SvIV(sv)
2946 # define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2947 # define SvUID(sv) SvUV(sv)
2948 # endif /* Uid_t_size */
2950 # if Gid_t_size > IVSIZE
2951 # define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2952 # define SvGID(sv) SvNV(sv)
2953 # elif Gid_t_sign <= 0
2954 # define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2955 # define SvGID(sv) SvIV(sv)
2957 # define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2958 # define SvGID(sv) SvUV(sv)
2959 # endif /* Gid_t_size */
2963 /* These are simple Marsaglia XOR-SHIFT RNG's for 64 and 32 bits. These
2964 * RNG's are of reasonable quality, very fast, and have the interesting
2965 * property that provided 'x' is non-zero they create a cycle of 2^32-1
2966 * or 2^64-1 "random" like numbers, with the exception of 0. Thus they
2967 * are very useful when you want an integer to "dance" in a random way,
2968 * but you also never want it to become 0 and thus false.
2970 * Obviously they leave x unchanged if it starts out as 0.
2972 * We have two variants just because that can be helpful in certain
2973 * places. There is no advantage to either, they are equally bad as each
2974 * other as far RNG's go. Sufficiently random for many purposes, but
2975 * insufficiently random for serious use as they fail important tests in
2976 * the Test01 BigCrush RNG test suite by L’Ecuyer and Simard. (Note
2977 * that Drand48 also fails BigCrush). The main point is they produce
2978 * different sequences and in places where we want some randomlike
2979 * behavior they are cheap and easy.
2981 * Marsaglia was one of the early researchers into RNG testing and wrote
2982 * the Diehard RNG test suite, which after his death become the
2983 * Dieharder RNG suite, and was generally supplanted by the Test01 suite
2984 * by L'Ecruyer and associates.
2986 * There are dozens of shift parameters that create a pseudo random ring
2987 * of integers 1..2^N-1, if you need a different sequence just read the
2988 * paper and select a set of parameters. In fact, simply reversing the
2989 * shift order from L/R/L to R/L/R should result in another valid
2990 * example, but read the paper before you do that.
2992 * PDF of the original paper:
2993 * https://www.jstatsoft.org/article/download/v008i14/916
2995 * https://en.wikipedia.org/wiki/Xorshift
2997 * https://www.iro.umontreal.ca/~lecuyer/myftp/papers/xorshift.pdf
2999 * http://simul.iro.umontreal.ca/testu01/tu01.html
3001 * https://en.wikipedia.org/wiki/Diehard_tests
3003 * https://webhome.phy.duke.edu/~rgb/General/rand_rate/rand_rate.abs
3007 /* 32 bit version */
3008 #define PERL_XORSHIFT32_A(x) \
3010 (x) ^= ((x) << 13); \
3011 (x) ^= ((x) >> 17); \
3012 (x) ^= ((x) << 5); \
3015 /* 64 bit version */
3016 #define PERL_XORSHIFT64_A(x) \
3018 (x) ^= ((x) << 13); \
3019 (x) ^= ((x) >> 7); \
3020 (x) ^= ((x) << 17); \
3023 /* 32 bit version */
3024 #define PERL_XORSHIFT32_B(x) \
3026 (x) ^= ((x) << 5); \
3027 (x) ^= ((x) >> 27); \
3028 (x) ^= ((x) << 8); \
3031 /* 64 bit version - currently this is unused,
3032 * it is provided here to complement the 32 bit _B
3033 * variant which IS used. */
3034 #define PERL_XORSHIFT64_B(x) \
3036 (x) ^= ((x) << 15); \
3037 (x) ^= ((x) >> 49); \
3038 (x) ^= ((x) << 26); \
3042 #endif /* PERL_HANDY_H_ */
3045 * ex: set ts=8 sts=4 sw=4 et: